The age of the opening of the Ice-Free Corridor and implications for the peopling of the Americas.

SignificanceThe Ice-Free Corridor (IFC) has long played a key role in hypotheses about the peopling of the Americas. Earlier assessments of its age suggested that the IFC was available for a Clovis-first migration, but subsequent developments now suggest a pre-Clovis occupation of the Americas that occurred before the opening of the IFC, thus supporting a Pacific coastal migration route instead. However, large uncertainties in existing ages from the IFC cannot preclude its availability as a route for the first migrations. Resolving this debate over migration route is important for addressing the questions of when and how the first Americans arrived. We report cosmogenic nuclide exposure ages that show that the final opening of the IFC occurred well after pre-Clovis occupation.

The “Ice-Free Corridor” (IFC), which developed as a contiguous ice-free area along the eastern front of the Rocky Mountains as the coalescent margins of the late-Pleistocene Cordilleran and Laurentide Ice Sheets separated and retreated (Fig. 1), has long played a central role in hypotheses about the peopling of the Americas (1–3). Upham (4) first proposed the existence of an IFC during the last glaciation, while Johnston (5) first proposed its use as a travel route from Beringia down to the Great Plains. The subsequent discovery of the Clovis cultural complex, which is dated as early as ∼13,400 calibrated 14C years ago (∼13.40 cal ka BP) (6, 7), was considered to be the oldest archaeological horizon in North America, forming the basis for the “Clovis-first” model for the migration south of the ice sheets that occurred by way of the IFC. At the same time, initial 14C ages from the IFC supported its availability as a migration route for Clovis people (8–10).

Fig. 1.

Map showing the extent of Cordilleran and Laurentide Ice Sheets after initial opening of IFC and location of dated sites. Margins of Cordilleran (CIS) and Laurentide (LIS) ice sheets at 14.2 cal ka are from ref. 55 and locations of sites with ages are discussed in Results. Sites with 14C ages are shown as blue and gray circles, with the former sites occurring more proximal to ice-margin retreat and opening of initial IFC or coastal corridor, and latter sites occurring more distal to initial IFC. Sites with luminescence ages are shown as orange and gray diamonds, with the former sites occurring more proximal to ice-margin retreat and opening of the IFC, and latter sites occurring more distal to the initial IFC. Sites with 10Be ages are shown as squares, with published ages are shown as blue squares constraining retreat of western Cordilleran Ice-Sheet margin (53, 54), green squares constraining opening of southern IFC (34), the black square constraining the age of deglaciation ∼500 km east of the IFC (41), and brown squares (with underlined text indicating site name) representing our mean ages that constrain the opening of the IFC. Inset map shows our sample locations within the context of the approximate location of the suture zone (thick dashed line) of the CIS and LIS at the LGM ∼21 ka. AB = Alberta, BC = British Columbia.

Several subsequent developments, however, have challenged the Clovis-first model as well as the corresponding role of the IFC in the peopling of the Americas. Challenges to the Clovis-first model have been largely driven by archeological (11–16) and genomic (17–19) studies that now provide compelling evidence for pre-Clovis occupation of the Americas south of the ice sheets by at least 15.5 to 16.0 ka (e.g., based on ages of ∼15.6 cal ka BP from the Cooper’s Ferry site [Idaho] (14) and estimates from genomic evidence of ∼15.7 ka [95% CI 17.5 to 14.6 ka] (19)). At the same time, assessments of newer age constraints have concluded that the IFC did not open until 14 to 15 ka (19–24). Moreover, ancient genomic and radiocarbon evidence indicate that the corridor only became suitable for human travel and subsistence (i.e., biologically viable) by ∼13.2 ka or earlier (25, 26), with other studies proposing even later IFC viability after ∼12.6 ka (27). These developments have lent support to a proposed migration route for pre-Clovis people from Beringia down the western Canadian coast (the “coastal corridor”) that largely bypassed the ice sheets (28–30). While some have suggested that the IFC route should not yet be rejected until the archaeological record supports the coastal corridor route (21, 31, 32), we note that no compelling archaeological evidence has yet been found to support a first migration through the IFC.

Despite broad agreement that the IFC was not unglaciated along its full length until 14 to 15 ka (19–24), existing ages used to support this time window are largely based on dating methods, many with large uncertainties, that include some lag between the time of deglaciation and the formation of the dated record (i.e., minimum-limiting ages) (26, 33). As we discuss further below, these same data can thus be used to argue for an earlier opening of the IFC that was available as a first-migration pathway. Similarly, arguments regarding the viability of the IFC to support first migrations are also based on dated organics that provide only minimum-limiting ages for biological productivity (25, 27). Uncertainties in existing data thus suggest that the IFC cannot yet be excluded as a potential route for pre-Clovis occupation. Resolving this debate over migration route is important for addressing the questions of when and how the first Americans arrived in regions south of the continental ice sheets.

To narrow these uncertainties in the current understanding of the age of the IFC opening, we first assess a compilation of ages that are commonly cited (in part or in entirety) to support the 14- to 15-ka time window for the age of the IFC opening (Figs. 1 and 2) (26). We then use 64 10Be surface exposure ages from 6 locations spanning ∼1,200 km of the Cordilleran–Laurentide suture zone that separated to produce the IFC (Fig. 1) to directly date the opening of the IFC with sufficient precision to establish whether it was available for the first peopling of the Americas south of the ice sheets.

Fig. 2.

Compilation of ages used to constrain the timing of the opening of the IFC. Sites with 14C ages are shown as blue and gray circles (2σ uncertainty), with the former sites occurring more proximal to Cordilleran–Laurentide ice-sheet suture zone and latter sites occurring more distal to ice-sheet suture zone (26) (). Sites with luminescence ages are shown as orange and gray diamonds (1σ uncertainty), with the former sites occurring more proximal to Cordilleran–Laurentide ice-sheet suture zone and latter sites occurring more distal to ice-sheet suture zone (26) (). Sites with 10Be ages are shown as squares (1σ uncertainty), with published ages as green squares constraining opening of southern IFC (34) (), gray (individual ages) and black (mean age) squares constraining the age of deglaciation ∼500 km east of the IFC (41), and our ages as brown squares (with corresponding site location names ABS-BCN) constraining opening of the IFC ().

Results

Previous Dating of the IFC.

Existing ages used to constrain the age of IFC opening include calibrated 14C ages that have been previously screened in order to remove those known to be potentially commonly contaminated and thus result in spurious results (e.g., bulk sediments, terrestrial shells) (26) (), a compilation of luminescence ages on sand dunes from Alberta (33) (), and cosmogenic nuclide (10Be) exposure ages on erratics from the southern section of the Cordilleran–Laurentide suture zone (34) (). We exclude four 10Be ages from this latter study in our analysis here because they are from sites immediately adjacent to two of our sampling sites, and so we instead include them with our new 10Be ages from those sites. We assess all ages according to their proximity to the Cordilleran–Laurentide suture zone (Fig. 1), with the expectation that they should become younger with increasing lateral distance from the ice-sheet suture zone.

14C ages typically provide only minimum-limiting ages on ice-margin retreat, owing to the unknown amount of time that lapsed between deglaciation and accumulation of reliably datable organic material (e.g., terrestrial plants) (35, 36). Of the 29 calibrated 14C dates from 7 sites that are within or proximal to the suture zone (Fig. 1), all but one are ≤13.5 cal ka BP in age, with no latitudinal trend (Fig. 2). Another 20 dates from 2 sites that are 200 to 400 km east of the suture zone are also ≤13.5 cal ka BP and show no longitudinal trend. Only one date is >13.5 cal ka BP (AA43652) and has been used to support the opening of the IFC by 15.0 ± 0.2 cal ka BP (32). This date, however, was measured on a standard collagen extraction from taiga vole bones that were redeposited by a low-energy mud flow. Another age on vole bones from the same unit are stratigraphically below but ∼1,600 y younger than the age 15.0 ± 0.2 cal ka BP. 14C ages on organics above and below the bone-bearing unit suggest the debris flow occurred during the middle Holocene (37). Given the potential for contamination of the collagen from old carbonate as well as the lack of stratigraphic integrity and provenance of the bones, this age provides a poor constraint on IFC opening (26). However, given their minimum-limiting nature, the remaining 48 dates do not preclude the opening of the IFC well before 13.5 cal ka BP. For example, other studies of Laurentide Ice-Sheet retreat that combined 10Be ages with minimum-limiting 14C ages found millennial-scale lags between the timing of ice retreat and the oldest 14C-dated material (36, 38, 39).

Luminescence ages on sand dunes in Alberta have been central to recent assessments of the age of the IFC (24, 26, 31–33), but these are even more problematic than the redeposited vole bone fragments. For example, the sampled sites cover a broad region of Alberta that extends from ∼150 km to 1,000 km east of the ice-sheet suture zone (Fig. 1), thus only providing minimum-limiting ages on the initial opening of the IFC. The fact that sediment sources for the sand dunes were largely derived from glacial lake sediments that became exposed following lake drainage (33) further limits their ability to date the initial opening of the IFC. Fig. 2 shows that ages from within relatively small regions can span 3.5 to 4.5 ky and that uncertainties on individual luminescence dates are high (mean ± 1 SE uncertainty is ±1.1 ky for 46 luminescence ages in ref. 33). In many cases, ages that are more proximal to the suture zone are younger than, or similar to, ages that are more distal, equivalent to a stratigraphic age reversal. Assuming full bleaching, the oldest luminescence age from the IFC zone should provide the closest constraint on ice-margin retreat. This would place the IFC as open before 15.3 ± 1.3 ka for the optically stimulated luminescence method (33) or 15.7 ± 1.6 ka for the infrared stimulated luminescence method (40).

The mean of 10Be exposure ages on erratics from sites extending between 49.1°N and 53.4°N has been used to argue that the southern sector of the IFC had opened by 14.9 ± 0.9 ka (14, 22, 24, 34). However, because these sites are spaced 10s to more than 200 km apart (Fig. 1), the ages of these erratics should not be averaged together to determine the millennial- to centennial-scale timing of the ice-margin retreat. We recalculated these ages following the same protocols as for our 10Be ages (Methods). Of the 11 dates considered here (excluding the 4 ages we combine with our 10Be ages [] and another age identified as a clear outlier because of its Holocene age), 5 sites have only one 10Be age and another three sites have only two 10Be ages, making outlier identification for each site difficult. If considered to contain no geologic scatter, these isolated 10Be ages would place IFC opening at 18.0 ± 0.8 ka (n = 1) at ∼49.1°N, 15.5 ± 0.9 ka (n = 2) at ∼50.3°N, and 17.6 ± 0.9 ka (n = 1) at 53.4°N; the remaining 8 ages fall within this range. While the ∼17.6-ka age appears to be out of stratigraphic order, one cannot exclude this age due to 10Be inheritance without a larger number of samples from the same site.

The mean of six 10Be ages from closely spaced samples was used to infer the formation of the IFC after 13.5 ± 0.8 ka (41). These samples, however, are ∼500 km east of the suture zone (Fig. 1), and their ages thus only suggest that the IFC opened before 13.5 ± 0.8 ka.

Cosmogenic Exposure Ages from the IFC.

To directly date the opening of the IFC with century-scale precision, we collected multiple samples for 10Be surface exposure dating from each of six sites () along a 1,200-km-long, south–north transect (∼50 to 59°N) where mapping identified the confluence of the Cordilleran and Laurentide Ice Sheets (42–44) (Fig. 1). These ages date the onset of ice-free conditions in association with ice-sheet separation and thus the initial opening of the IFC at that location. To exclude the possibility that samples were exhumed from moraine degradation, we sampled glacially transported boulders resting on bedrock and, at site British Columbia north (BCN), glacially eroded bedrock (Fig. 3). Five of the sites are on topographic highs above the elevations of former proglacial lakes (45), and there is no evidence of lake cover at the one site on the plains of west–central Alberta (ABC) (34, 43). We collected >10 samples at 5 of the sites and 5 samples from a 6 site (British Columbia central [BCC]) to assess geologic scatter. Given their proximity to our two southernmost sites (Alberta south [ABS], ABC), we also include four 10Be ages from ref. 34. (Fig. 1 and ).

Fig. 3.

Photographs of representative glacial erratics sampled for cosmogenic 10Be dating. (A) Sample ABC-5-16. (B) Sample ABS-1-16.

We calculated individual 10Be ages using the Arctic production rate (46) and Lal/Stone time-varying scaling (47). We use the Arctic production rate calibration data set (46) for our high-latitude sites as opposed to the global production rate dataset (47) because it better accounts for the regional variability of the cosmic ray flux and associated production rate. We note, however, that the exposure ages calculated using the global production rate calibration data set are, on average, only 4% younger than our preferred exposure ages, which is within the 1σ external uncertainties.

We made no snow cover, vegetation, or erosion corrections. Given the major climate shifts of the last 14,000 y, it is impossible to know the snow-cover history of the sites, but several factors mitigate the effect of snow cover, as follows: snow has a low density and thus low shielding of cosmic rays; we sampled from the tops of boulders, where snow cover is more likely to blow off by wind; and we find excellent agreement between boulders and bedrock samples at site BCN where the former should have a lesser impact from snow cover than the latter. Vegetation similarly changed over the last 14,000 y, but as with snow cover, it has a low density and thus low shielding of cosmic rays. We sampled boulder surfaces that showed little-to-no erosion, but typical erosion rates (<1 mm ka−1) suggest that erosion is not an issue at these timescales (48). We accounted for changes in sample elevation from isostatic uplift by using an iterative approach with isostatic rebound models (49) and ice-sheet simulations (50). After removing outliers based on the constraints that samples cannot date from older than the Last Glacial Maximum (LGM; 26.5 to 19 ka) or be from the Holocene (<11.7 ka), we calculated the timing and uncertainty of deglaciation at each site using the arithmetic mean and SE (Fig. 2 and ), with the latter also including production rate uncertainty and reported as the external uncertainty to facilitate comparison with other dating methods for IFC opening.

Our 10Be ages suggest that the IFC opened first at 15.4 ± 0.7 ka at our southernmost site (ABS, 50.0°N, n = 9). The IFC then progressively opened from the north and the south, with our mean age of 14.8 ± 0.7 ka at site ABC (52.2°N, n = 10) being (within uncertainty) the same as 10Be ages from our two northern sites that suggest the northern section of the IFC began to open (14.4 ± 0.6 ka, BCN, 58.7°N, n = 10; 14.5 ± 0.7 ka, BCC, 56.1°N, n = 4) (Fig. 2). The center of the IFC then opened at 14.2 ± 0.6 ka at site Alberta north (ABN) (53.9°N, n = 11) and lastly at 13.8 ± 0.5 ka at site British Columbia south (BCS) (56.1°N) (Fig. 2), with the latter mean age being particularly robust as it is based on 15 10Be ages with no outliers. Indeed, the SE at this site is only ±0.1 ka (±0.7%), with the remaining ±0.4 ka coming from the production rate uncertainty of ±3.8% (46). A Monte Carlo experiment using the distribution of 10Be exposure ages from the BCS site suggests that ∼>50 exposure ages are required to drive the site SE below ±0.1 ky (), indicating that our analyses for this site are at the practical limit of precision for this direct dating technique and the ±0.5-ka external uncertainty cannot be realistically improved upon without reduction in production rate uncertainty. Finally, we note that site BCS (56.1°N, 122.2°W) is located just west of two sites (Charlie Lake, 56.3°N, 120.9°W; Spring Lake, 55.5°N, 119.6°W) where evidence of steppe vegetation and a variety of animals first appears at ∼12.6 ka (27), suggesting that the final opening of the IFC occurred ∼1.2 ky before this area became biologically viable.

Fig. 2 compares our 10Be ages with the existing chronology for the IFC assessed above. Except for the calibrated 14C age on the taiga vole bones (AA43652), all calibrated 14C ages are younger than the mean 10Be ages, consistent with the 14C ages being minimum-limiting ages on deglaciation. At the same time, the calibrated 14C age on taiga vole bones is ∼1 ky older than the mean 10Be age of 13.8 ± 0.5 ka from the nearby BCS site and ∼2.2 ky older than nearby sites recording the first colonization of the area by plants and animals (27), which is consistent with the bones being contaminated by old carbonate (26). Many luminescence ages that are proximal to, but still east of, the central IFC area are older than the mean 10Be ages for deglaciation from these latitudes, further identifying age reversals that are possibly due to mixing with incompletely bleached glaciofluvial or glaciolacustrine sand prior to burial. Published 10Be ages from the southern IFC include two that are significantly older than our mean ages and one (8.6 ± 0.3 ka) that is significantly younger (34). The remaining ages are similar to our mean ages, but again cannot be combined as a single population to calculate a mean age because of the large distances between the sites. Of the two published ages included with our ABS ages, one is an outlier (181 ± 4 ka) and the other (15.8 ± 0.5 ka) is similar to the mean age (15.4 ± 0.7 ka). Of the two published ages included with our ABC ages, one is also an outlier (24.2 ± 0.9 ka), whereas the other age (16.4 ± 0.6 ka) is significantly older than the mean age (14.8 ± 0.7 ka), providing an example of how a single age affected by geologic scatter can provide an inaccurate age for IFC opening. Finally, the 10Be ages from the Beaver River site in west–central Saskatchewan ∼500 km east of the suture zone (mean age of 13.5 ± 0.8 ka) (41) are consistent with the ∼13.8-ka timing for opening of the IFC from our 15 10Be ages at site BCS.

Discussion

Our 10Be chronology closely dates the final opening of the IFC as occurring at 13.8 ± 0.5 ka (Fig. 4), establishing that the IFC was not available as a migration route for the peopling of the Americas that had occurred before ∼15.6 ka suggested by current archaeological and ancient genomic evidence (Fig. 4). In contrast, multiple 14C and 10Be ages (51–55) from the coastal corridor () indicate that retreat of the western margin of the Cordilleran Ice Sheet and associated postglacial uplift occurred early enough to have provided a migration route for people prior to the earliest known archaeological sites south of the ice sheets (Fig. 4). We note that these first peoples would likely still have faced considerable difficulties in navigating the largely glaciated coastline (51). Further dating of western CIS margin retreat and relative sea level change will improve our understanding of the paleogeography of this coastal corridor, which can then serve as a guide in the search for the archaeological evidence that is required to confirm this coastal migration route.

Fig. 4.

Age constraints on opening of IFC, coastal corridor, and first peopling of the Americas south of the ice sheets. (A) Sites with 14C (2σ uncertainty) and 10Be (1σ uncertainty) ages constraining retreat of western margin of the Cordilleran Ice Sheet and emergence of ice-free areas are shown as blue horizontal lines and squares, respectively (52–55) (). Sites with 10Be ages constraining opening of the IFC are shown as brown squares (1σ uncertainty) (this study). (B) Age constraints from ancient genomics (19) and archaeological sites for first peopling of the Americas south of the ice sheets. Ages from archaeological sites () are labeled as Paisley Caves, OR (PC) (12); the Page-Ladson site, FL (PL) (15); the Debra L. Friedkin site, TX (DF) (16); and the Cooper’s Ferry site, ID (CF) (14).

Methods

Field Sampling.

We collected multiple samples for 10Be surface exposure dating from each of six sites located along a 1,200-km-long south–north transect (∼50 to 59°N) () where mapping identifies the confluence of the Cordilleran and Laurentide Ice Sheets (42, 43, 56) (Fig. 1). The 10Be ages at each site date the onset of ice-free conditions from ice-sheet separation and thus the initial opening of the IFC at that location. This contrasts with existing geochronological data used to constrain the age of IFC opening which only minimum-limiting ages on ice-margin retreat since they record some unknown amount of time between deglaciation and accumulation of reliably datable material as well as the fact that many dated sites extend from ∼150 km to 1,000 km east of the ice-sheet suture zone.

Samples for 10Be surface exposure ages from boulders on bedrock were collected at five sites (ABS, ABC, ABN, BCS, BCC) with a sixth site (BCN) consisting of a mixture of boulder-on-bedrock samples (n = 4) and glacier-scoured bedrock samples (n = 6). Sampling followed prior methods used to document ice-sheet-scale changes in ice margins and area of the IFC (4, 5, 38, 57). Each of these sites is located near the mapped confluence of the Cordilleran–Laurentide Ice Sheets (Fig. 1) and was chosen for its simple deglacial history, with the exposure of the boulder surface being due to separation of the two ice sheets with no potential cover by proglacial lakes.

ABS is the southernmost site (50.1°N, 113.8°W, ∼1,200 m above modern sea level); we collected 10 samples from this sample site and include 2 additional ages from ref. 34. (ALT-MM-15-08, ALT-MM-15-09) (). All boulders are quartzite except ABS-10-16 which is granitic. At the ABC site (52.2°N, 114.8°W, ∼1,100 m above modern sea level), we collected samples from 12 quartzite boulders and combined them with 2 ages from ref. 34. (ALT-MM-15-14, ALT-MM-15-15) (). ABN (54.0°N, 119.0°W, ∼1,800 m above modern sea level) consists of 12 quartz-rich sandstone boulders (). At BCS (56.1°N, 122.2°W, ∼1,100 m above modern sea level), we sampled 15 sandstone boulders resting on bedrock (), while we found only 5 samples for collection at BCC (57.5°N, 122.9°W, ∼1,200 m above modern sea level) (). For BCN (58.7°N, 123.8°W, ∼1,000 m above modern sea level), we sampled only 4 sandstone boulders (BCN-4-16, BCN-5-16, BCN-7-16, BCN-8-16), which we supplemented with an additional 6 sandstone bedrock samples to give a total of 10 samples ().

10Be Target Preparation and Measurement.

Thirty-four samples were prepared at PRIME Laboratory at Purdue University (6 = ABS, 8 = ABC, 5 = ABN, 5 = BCS, 5 = BCC, 5 = BCN) following standard techniques (www.physics.purdue.edu/primelab/) with another 30 samples prepared at CosmIC Laboratories at Imperial College London (4 = ABS, 4 = ABC, 7 = ABN, 10 = BCS, 0 = BCC, 5 = BCN) using standard procedures (8). 10Be concentrations and uncertainties are provided in .

Age Calculation and Isostatic Rebound.

We calculated 10Be ages using the Lal/Stone time-varying scaling (47) and the Arctic 10Be production rate (46). Use of another scaling (LSD; 58) does not change our results or conclusions. We do not correct for changes in atmospheric pressure as this has been shown to have a negligible impact on the production rate once the total time since deglaciation is considered (38, 57). Because the Cordilleran–Laurentide Ice-Sheet suture zone has undergone extensive isostatic rebound following deglaciation (49, 50), we estimate the impact of rebound on the surface exposure ages and correct for this influence on the final age (59). This is accomplished using ice-sheet model simulations following ref. 50.

We calculated average deglacial ages for each site following the methods of refs. 38 and 57 and include the four additional 10Be ages from ref. 34 that are adjacent to ABS and ABC (). Outliers that are either at least as old as the LGM (>19 ka) or Holocene (<11.7 ka) in age were identified and excluded, which excluded - samples from the 68 samples, 2 of which are from ref. 34. ABS contained three pre-LGM outliers while ABC had four LGM-age outliers (). One sample at ABN and another at BCC were identified as Holocene-age outliers (). Following exclusion of these nine outlier samples, we determined the arithmetic mean and SE for each site, which produced the largest uncertainty in the mean, making this a conservative approach (38). We then added the 10Be production rate uncertainty of ±3.8% to the SE in quadrature to determine the external uncertainty of the mean for the timing of deglaciation and the opening of the IFC.