The Smoking Hills area, in the western Canadian Arctic (~69°15’N; 127°00’W), is a Cretaceous bedrock upland (~380 m asl) that has long been considered to have formed a nunatak during the Late Wisconsinan Laurentide glaciation, encircled by the westward-flowing Amundsen Gulf ice stream and northward flowing ice from the Great Bear Lake divide.

Recent investigations of the glacial history on Banks Island, to the immediate north of Smoking Hills, have demonstrated its inundation by the Late Wisconsinan Laurentide Ice Sheet (England et al., 2009; Lakeman and England, 2013). Stratigraphic sections on Banks Island at Duck Hawk Bluffs, Worth Point, and elsewhere, purported to contain multiple glaciations, interglacials and preglacial fluvial deposits (Vincent, 1983, 1990), have been fundamentally revised, and now indicate perhaps only 2 glaciations (Evans et al., 2014; Vaughan et al., 2014). Of relevance to our Smoking Hills study, these stratigraphic re-assessments also demonstrated widespread glacitectonism, including large rafts of intact, poorly consolidated Cretaceous bedrock, which had not previously been recognized.

In the Smoking Hills in 1968, Rudy Klassen discovered a 57 m high Quaternary section in a small catchment west of the lower Horton River in which he identified a basal sand and gravel deposit, overlain by 3-5 tills containing various inter-till/interglacial deposits. Vincent returned to this section in 1988 and 1991, and described at least 3 tills separated by lacustrine, interglacial, and possible paleosol deposits, and undertook preliminary paleomagnetic measurements that indicated a lower magnetic normal till (? Gauss Chron; 3.04-2.58 Ma), overlain by reversed and then normal magnetic tills (? Matuyama (2.58-0.78 Ma) and Brunhes chrons (0.78-0 Ma), respectively). Duk-Rodkin and Barendregt returned to this site in 2004 and conducted extensive magnetostratigraphic sampling (Barendregt and Duk-Rodkin, 2004; Duk-Rodkin and Barendregt, 2011). They indicated that the basal sand and gravel deposits (what they termed “pre-glacial” fluvial gravels) were reversely magnetized (? Matuyama Chron), and that these were capped by 3 magnetically reversed tills, separated by lacustrine/interglacial deposits, which was then capped by normally magnetized (? Brunhes Chron) interglacial deposits and an uppermost normally magnetized till. They indicated that Canadian Shield erratics were absent from the entire stratigraphy, other than a scattering of granite boulders on the top of the section. From this, they interpreted that all the underlying tills were the product of a local “Horton Ice Cap” source, and that a continental Laurentide glaciation was only responsible for depositing the uppermost surface lag (age unknown).

In 2018, during 2 weeks of helicopter-supported fieldwork in the Smoking Hills, we studied Klassen’s stratigraphic section in detail, and discovered several other notable sections in the area. Detailed sedimentological, structural, lithological, cosmogenic burial dating, and palynological analysis of these sections were undertaken, along with a glacial landsystem approach to mapping and reconstruction of the glacial geomorphology. Based on our results and observations, we have developed a fundamentally different understanding of the glacial history of the region that extends the glacial record back to late Pliocene and records a Late Wisconsinan Laurentide Ice Sheet inundation of the region (Evans et al., accepted).

Numerous observations made by us at Klassen’s stratigraphic section change its interpretation. The basal sand and gravel lithofacies contains both faceted and striated clasts, and has a lithological composition that for the most part could not have been derived from the local Cretaceous bedrock catchment. We interpret this basal unit as a glacifluvial deposit, or at least the fluvial reworking of proximal glacial sediments. A reported ice wedge pseudomorph within these sediments (unobserved by us due to slumping; but another example was seen in a different section that we believe to be contemporaneous glacilacustrine sediments) suggests either an earliest glacial, or an oscillating ice margin that subsequently deposited the overlying diamict stratigraphy. A cosmogenic burial age of ~2.9 Ma (Evans et al., accepted) from the lowermost diamict lithofacies places this within the latest Pliocene, roughly coeval (within its standard error) with the most extensive Cordilleran Ice Sheet glaciation recorded in Yukon (2.64 Ma; Hidy et al., 2013).

The diamict sequence overlying the basal glacifluvial gravel is unlike what has been described previously. We recorded three lithofacies comprising massive to laminated, clast-poor diamict (2-10, 16-44, and 47-51 m depth), containing prominent rafts, intraclasts, boudins and stringers of coherent to highly glacitectonised poorly consolidated Cretaceous bedrock (confirmed by palynology and sedimentology). These include two 6 m and 3 m thick rafts of bedrock. Prior studies did not identify any of these, and instead have confused them, we believe, for lacustrine, mudflow, and other deposits (Duk-Rodkin and Barendregt, 2011). Glacitectonism of Cretaceous bedrock, in cases extending up to 20 m depth, was identified by us as being widespread in the field area, and explains some of the overly thick stratigraphic sequences we found. Detail is not provided on what sediments in the Klassen section were sampled by past paleomagnetism studies, but comparisons of photos in Duk-Rodkin and Barendregt (2011), and a tendency for such studies to target fine sediments, leads us to suspect that many of the samples could have been taken from Cretaceous bedrock rafts. We are thus inclined to disregard the published magnetostratigraphy, not least because the reversely magnetized characterization of the lowermost diamict lithofacies contradicts our cosmogenic burial age, which should instead fall within the Gauss Chron normal.

Canadian Shield-derived granitic clasts were observed by us (and also by Vincent) within all of the diamict deposits at the Klassen section, but not in the basal glacifluvial gravel. This contradicts the previous notion that these deposits were derived from a “Horton Ice Cap”, situated north of the Canadian Shield. Instead, it argues for continental Laurentide Ice Sheet sources, and indeed, despite their absence, we do not rule out that the basal glacifluvial gravels could also be produced by the first Laurentide glaciation of the region.

Clast fabrics from the diamict lithofacies record flow generally from the east and southeast, which we relate to glacially streamlined landforms, geomorphology, and ice marginal landform assemblages accordant with past ice flows emanating from Amundsen Gulf and Great Bear Lake sectors of the Laurentide Ice Sheet. The uppermost diamict facies (2-10 m depth) is broadly similar to those underlying it, but exhibits sufficient chemical, sedimentological, and colour variation that we consider it distinct from the underlying stratigraphy, and instead relate it to the last (Wisconsinan) glaciation. In the absence of chronological control, we construct a deglacial landsystem model using surface geomorphology, regional ice flowsets (e.g., Stokes et al., 2006), controlled moraines, and ice marginal erosional and depositional features to illustrate a coherent uncoupling of formerly confluent ice, that inundated the entire landscape, into separate ice lobes retreating eastward into Amundsen Gulf, southeastward towards Great Bear Lake, and southward along the Mackenzie River valley.

One site we discovered along the Anderson River exposes what may be a unique occurrence of buried (pre-last glaciation) glacial ice. A 250 m wide retrogressive thaw slump, exposes an 8 m high headwall of clast-rich, foliated glacier ice, which we presume had stagnated and then became buried by a carapace of insulating sediments. As permafrost aggraded into the site, large vertical ice-wedges formed within the buried glacial ice. Subsequently, the upper extents of these ice wedges were melted, producing a prominent thermal unconformity, and then, or coeval with, they were deformed obliquely downslope, parallel to what we identify as adjacent ice-cored, pitted flutings. Evidence suggests that both the thermal unconformity and deformation of the ice wedges occurred subglacially, as the overriding ice became coupled with the underlying relict glacial ice, forming a glacitectonite shearing interface. The implications of this site are profound for both the historical stability of permafrost conditions and the subglacial dynamics likely to have occurred in areas of northern Canada hosting extensive relict glacier ice, overridden by later glaciations.

The Smoking Hills was found to contain a rich sedimentological record of 2 or more past glaciations, unique for mainland arctic Canada (outside of the Cordillera). Glacial modelling which has relied on past magnetostratigraphies and local ice cap versus continental ice sheet Quaternary glacial histories in this region, require re-examination. We are endeavoring to return to this site to conduct follow up detailed studies of the buried glacier ice. We also will study well preserved and abundant intra- and sub-till wood, peat and other organic macrofossils (including what appears to be a glacial rafted overbank deposit with rooted trees, leaf litter mats and marl), and to conduct OSL and further cosmogenic dating, as a test of our results and to further constrain the basal gravel sequence and other deposits.