Unravelling the fragmented sediment–landform assemblage in an area of thick Quaternary sediment, western Hudson Bay Lowland, Canada

Regions of thick Quaternary sediment cover, such as the western Hudson Bay Lowland (HBL), provide an important sedimentary record of ice-sheet changes over time. The Quaternary stratigraphic record can enable an understanding of ice-sheet growth and decay over multiple glacial–interglacial cycles, beyond what can be gleaned from the surficial landscape, and provide important insights into past interglacial climates (Nielsen et al. 1986; Dredge et al. 1990; Gao et al. 2012; Hodder et al. 2016; Norris et al. 2018; Möller et al. 2019; Evans et al. 2021). At the surface, interior regions of paleo-ice sheets consist of fragmented patches of differently-oriented streamlined landforms and outcrop-scale ice-flow indicators, with palimpsest sediment dispersal patterns (Dyke et al. 1982; Prest 1990; McMartin and Henderson 2004; Veillette 2004; Gauthier et al. 2019; Rice et al. 2019; McMartin et al. 2021). Such complex landscape records present many challenges yet provide an opportunity to better understand the long-term behavior of ice sheets (Dalton et al. 2022a). To advance reconstructions of pre-last glacial maximum (LGM) ice sheet dynamics, beyond what can be understood from the landscape record, the palimpsest stratigraphic record in the relatively rare regions that contain these records needs to be deciphered.

The Laurentide Ice Sheet (LIS) overran the western HBL during the last glaciation, with ice emanating from both the Keewatin and Quebec–Labrador (Q–L) domes (Shilts 1980; Prest 1990), and either a separate “Hudson dome” (Dyke et al. 1982) or late deglacial Hudson Bay Ice Saddle (Dyke and Dredge 1989). How these competing ice centres interacted is mainly understood from the LGM through to deglaciation (Dalton et al. 2023), but less so for glacial events prior to that (Kleman et al. 2010; Batchelor et al. 2019). The Quaternary stratigraphy of the HBL is the net result of erosion and deposition that spans at least three glacial-interglacial cycles (Nielsen et al. 1986; Thorleifson et al. 1992; Dredge and McMartin 2011). Regional reconstructions of the relative ice-flow history have identified evidence for ice flowing toward the northwest and/or west (Q–L dome) during both pre-LGM growth and a previous glacial cycle (Gauthier et al. 2019). Understanding the extent, relative timing, and number of northwest-oriented ice-flow phases in the western HBL is important because they must have occurred at times when the Keewatin dome was smaller or situated further to the west compared to its deglaciation position.

The geomorphic record of the western HBL is composed of a complex and fragmented landscape (Gauthier et al. 2019). This is exemplified in the study area, where two streamlined-landform flowsets are aligned at near perpendicular angles and situated within 5 km of each other (Figs. 1 and 2), but do not crosscut (Gauthier et al. 2019). The larger flowset represents south-southwest-trending ice flow and is interpreted to have been formed during deglacial ice streaming by the Hayes Lobe (Figs. 1 and 2; Dredge and Cowan 1989; Margold et al. 2018; Gauthier et al. 2019). Here, we focus on the isolated Kaskattama flowset that occurs on a relatively flat-topped hill, informally referred to as the “Kaskattama highland”. The timing and direction of the northwest–southeast oriented Kaskattama flowset is unknown (Figs. 1 and 2; Gauthier et al. 2019). The “highland” contains thick Quaternary-aged sediments (Hodder 2018; Hodder and Gauthier 2021), and forms a raised (∼130 m above the surrounding lowland; Fig. 1) and isolated landscape whose surface geomorphology does not fit with the surrounding deglacial ice-flow record. Unravelling this poorly understood part of the Quaternary stratigraphic record in the western HBL will help advance regional ice-sheet reconstructions and understanding. To do this, new stratigraphic strategies that take into account the fragmented nature of the sediment–landform assemblages (as described in Gauthier et al. 2019) were developed.

Quaternary sediment cover is extensive in the study area and the only known bedrock outcrops are carbonates within the bed of the Gods River, near the First Nation community of Shamattawa (Fig. 1). The Gods River has incised through 15 to 30 m of Quaternary sediments at this location, but less is known about the thickness of Quaternary sediments elsewhere in the study area and throughout most of the western HBL region (Hodder and Gauthier 2021). Near Shamattawa, a prominent drainage corridor formed during the Holocene when this area was an eastern outlet for Lake Agassiz (Dredge 1983). Here, the bottom of this drainage corridor is at an elevation of ∼80–84 m above sea level (asl) and the Hayes flowset on either side of the corridor is at an elevation of ∼125–144 m asl, indicating that ∼45 to 60 m of the stratigraphic record was eroded (Fig. 2). The limit of postglacial marine transgression around the highland is ∼145 m asl (Gauthier et al. 2020), and the geomorphic record below this limit is partially or completely covered by marine sediments.

The study area is primarily underlain by carbonate Paleozoic rocks (limestone and dolomite) of the Hudson Bay Basin (HBB; Manitoba Geological Survey 2022), which unconformably overlie the Precambrian Canadian Shield surround the HBB (Fig. 3A). Rocks of the Precambrian Canadian Shield are composed primarily of granite and gneiss with scattered greenstone belts.

In northern Manitoba, distinctive erratics indicative of continental-scale dispersal include both northwestern- and eastern-sourced lithologies. Distinctive erratics from the Dubawnt Supergroup have been transported toward the southeast over 800 km (Fig. 3A; Rainbird et al. 2003). Examples include transported clasts of the Pitz Formation (Fig. 3B; purple to mauve-coloured volcanic rocks with distinctive chalky sanidine phenocrysts) and Christopher Island Formation (red volcanic rocks with phlogopite phenocrysts). Eastern erratics that have been transported toward the west, a minimum of 650 km, include the Belcher Group (Fig. 3A) greywacke with hemispherical calcareous concretions (Fig. 3C) that are sourced from the Omarolluk Formation (“Omars”; Prest 1990; Prest et al. 2000; Jackson 2013) and oolitic jasper derived from the Kipalu Formation (Fig. 3D). Oolitic jasper can also be sourced the Sutton Inlier situated 400 km to the southeast, where metabasalt also outcrops (Fig. 3A; Stott et al. 2010).

Streamlined landforms from two separate ice-flow events overlie parts of the study area—grouped and named as the Hayes and Kaskattama flowsets (Fig. 2; flowsets F and K, respectively, of Gauthier et al. 2019). The south-southwest-trending Hayes flowset (204°–207° in the study area) is a large (400 by 300 km) flowset that extends beyond the study area to the southwest and is interpreted to have been formed by the Hayes Lobe ice stream (Dredge and Cowan 1989; Margold et al. 2018; Gauthier et al. 2019, 2022). The Kaskattama flowset is smaller (85 × 60 km) and is oriented northwest–southeast, extending across the Kaskattama highland at a mean elevation of 165 m asl. It is characterized by curvilinear glacially streamlined landforms (Fig. 2) with orientations varying from about 110°-290° on the east side of the Kaskattama highland to 125°-305° at the northwest edge the Kaskattama highland. Streamlined-landform elongation ratios are between 4.5:1 and 10:1. The Kaskattama flowset is an isochronous event flowset (Greenwood and Clark 2009; Gauthier et al. 2019). This is because the flowset has high parallel conformity but is isolated from the surrounding geomorphology. It is transversely crosscut by meltwater corridors parallel to the Hayes Lobe (Fig. 2), which indicates that the event occurred prior to the formation of the Hayes flowset (Gauthier et al. 2019). The boundary between the two streamlined-landform flowsets is one of the meltwater corridors, and individual landforms do not meet or crosscut. The terrain on the north edge of the Kaskattama highland is hummocky and lacks streamlined landforms at the 30 m resolution of available elevation models (Shuttle Radar Topography Mission; United States Geological Survey 2014).

There is an exceptional, yet fragmented, depositional record of intertill nonglacial deposits in the western HBL (McDonald 1969; Netterville 1974; Shilts 1984; Nielsen et al. 1986; Dredge et al. 1990; Dalton et al. 2016), deposited during at least two pre-Holocene interglaciations (Dredge and McMartin 2011). Two previously-studied organic-bearing interglacial beds are located in the study area, along tributaries of the Echoing River (McDonald 1969; Dredge et al. 1990) and along Gods River (Netterville 1974; Klassen 1986; Mesich et al. 2023). These sediments, named after their rivers, both contain wood that have nonfinite radiocarbon ages and paleoenvironment records indicating climate conditions similar to the present day (Dredge et al. 1990). The age of these subtill interglacial beds is unknown but based on stratigraphic relationships they are presumed to be ∼Marine Isotope Stage 5 (MIS 5; Dredge et al. 1990; Dredge and McMartin 2011). Furthermore, preliminary stratigraphic investigations along the Gods River indicate at least two separate subtill interglacial beds are preserved in the stratigraphic record exposed (Mesich et al. 2023).

To investigate the stratigraphic record of the Kaskattama highland and surrounding area, helicopter-supported fieldwork was undertaken in 2016, 2017, and 2019. Surficial sediments were characterized from hand-excavated pits and/or using a Dutch auger (Figs. 4A and 4B), while subsurface sediments were described from natural sections exposed along rivers and meltwater corridors. Sections were first cleared to remove colluvium, then the in situ sediments were described over a 1–2 m wide interval based on their texture, colour, relative consolidation, clast content, sedimentary structure, and nature of lower contact. Thick colluvium inhibited a thorough assessment of the lateral continuity of most identified sedimentary packages. Samples of surficial diamicton interpreted to be subglacial till were collected from the C-horizon. At sections, till samples were collected above and below all facies changes in the field (colour, texture, or change in compaction) and from every 1–4 m even when visible stratigraphic changes could not be observed. Stratigraphic observations are summarized on stratigraphic columns using standard lithofacies codes (Evans and Benn 2021). Detailed sediment descriptions, coordinates, and photos for all sections cited in the text (e.g., Section 16-416) are provided in Hodder et al. (2024) and key sections not presented within the manuscript are provided in supplementary material A.

To characterize the ice-flow direction during till formation, indicators of ice flow were measured from sections investigated. Till beds observed in the study area are generally massive and no glaciotectonic features were observed to supplement or contradict ice-flow interpretations. Ice-flow data include clast fabrics (e.g., Figs. 4D–4F), as well as the orientations of the youngest parallel striations on the upper surfaces of in situ, lodged, cobble- to boulder-sized clasts, which are considered a good indicator of paleo-ice flow direction (Fig. 4G; Stokes et al. 2008; McMartin and Paulen 2009). Clast fabrics provide information on the direction of shear during till deposition (e.g., Hicock et al. 1996; Lian and Hicock 2000) and determining this direction was the sole purpose of measuring clast fabrics. Clast-fabric measurements were conducted on three vertical faces of a “box” carved into the section. At each fabric site, the trend and plunge of stone a-axes were measured from at least 30 clasts where the a:b axis ratio was >1.5, the a-axis plunge was <70° and the b-axis plunge was <60°. At one site only 20 clasts were measured due to a field time constraint. A nonmagnetic knitting needle was used to extrapolate the plunge of the a-axis out of the till face, allowing for increased accuracy of measurements. We did not measure clasts if they were in contact with other clasts or smaller clasts that were adjacent to larger clasts, which could restrict their rotation and alignment as passive markers of strain (Andrews and Smith 1969; Evans et al. 2016, 2018). The modality of each fabric was assessed visually (Hicock et al. 1996) and eigenvectors and eigenvalues were calculated (Mark 1973; Benn 1994) using Rockware StereoStat v1.6.1. This was a necessary step to identify spread-bimodal or multi-modal fabrics where no ice-flow direction could be confidently interpreted. All ice-flow-orientation interpretations based on clast-fabrics were initially considered bi-directional, since the up-ice plunge is not consistent (Andrews and Smith 1969; Saarnisto and Peltoniemi 1984; Larsen and Piotrowski 2003; Gauthier et al. 2019). In these cases, other factors such as regional ice-flow history and/or provenance of clasts within the till may provide an indication of ice-flow direction (e.g., Catto 1998). Existing studies in the western HBL do not document north-, northeast- or east-trending ice flow (Gauthier et al. 2019), so all similarly-oriented clast fabrics are interpreted to have formed during south-, southwest-, or west-trending ice flow, respectively. The final interpretation is further supported by till provenance analysis.

In northern Manitoba both northwest- and southeast-trending ice flow is documented in the stratigraphic and erosional ice-flow records (Gauthier et al. 2019). This makes the interpretation of ice-flow indicator data, such as clast fabrics from till and the northwest–southeast oriented Kaskattama flowset, complicated. Till-provenance analysis is used in this study to complement the ice-flow orientation data for each identified stratigraphic bed. Provenance analysis is a group of techniques that are used together to investigate the composition of sediment to infer their bedrock source(s). It is thus used to solve an “inverse problem”, whereby the composition of sediment allows one to determine the nature of the bedrock source area (e.g., Lipp et al. 2021). A common approach for poorly-sorted sediment like tills is to investigate the lithology of the coarse fractions, typically referred to as “clast lithology counts” (e.g., Shilts 1980). This is often complemented with an analysis of sand-sized indicator minerals and matrix geochemistry (e.g., McClenaghan et al. 2022). To discriminate among possible bedrock sources, dating the formation or metamorphic age of mineral grains (e.g., zircon or hornblende) is another approach that has been applied on both glacial erratics (e.g., Doornbos et al. 2009) and smaller detrital grains in tills (e.g., Roy et al. 2007a; Kassab et al. 2017) or ice-rafted debris layers (Hemming 2004; Roy et al. 2007b). In this study, till clast-lithology counts and detrital hornblende ⁴⁰Ar/³⁹Ar ages are used together to constrain the bedrock sources of till detritus. Matrix geochemistry is used to further support these interpretations.

The clast lithology types identified within the 2–8 mm size-fraction of till were first sorted into 16 detailed classes and then grouped into three simplified classes reflective of the regional bedrock geology: HBB, granitoid and undifferentiated greenstone and greywacke (UGG; Hodder et al. 2024). Interestingly, there is a background concentration of ∼10% granitoid clasts within till sampled across the HBB, as noted by Shilts (1980). Hence, they are not an effective indicator to delineate dispersal patterns at low concentrations. To resolve the northwest–southeast direction problem, we focus on the UGG class (e.g., Fig. 3E). The source of these indiscriminate fine-grained, dark grey rock fragments is presumed to be the Belcher Group to the east, or Sutton inliers to the southeast (Fig. 3A; Shilts 1980; Thorleifson et al. 1993). Thus, tills with elevated UGG concentration are interpreted to have been derived from the east and deposited by ice flowing from the Q–L sector of LIS. Similarly, oolitic jasper clasts are found in trace concentrations in till in the study area (e.g., Fig. 3D) and have a presumed eastern to southeastern provenance (Fig. 3A; Belcher Group and Sutton inlier). Dubawnt Supergroup erratics, sourced to the northwest of the study area, are found in trace concentrations in till within the study area (Fig. 3B). An average of 377 clasts were counted for each till sample, ranging from 224 to 540 clasts per sample. Two surficial till samples have anomalous proportions of Precambrian clasts (62.7 and 79.4 ct. %; supplemental material C) and were removed from the dataset prior to analysis. The potential significance of these results is discussed in Hodder et al. (2024).

The ⁴⁰Ar/³⁹Ar age of hornblende grains records time elapsed since they were last subject to temperatures exceeding the closure temperature of the mineral, which for hornblende grains is approximately 510 °C (Schaen et al. 2021). The Canadian Shield is Precambrian in age and categorized into geological provinces based on the tectonic history of these regions (e.g., Hoffman 1989). Three geological provinces are important to this study in terms of the detrital hornblende ⁴⁰Ar/³⁹Ar ages: Superior, Western Churchill (WC), and Trans Hudson Orogen (THO) provinces (Fig. 3A), which have approximate hornblende ⁴⁰Ar/³⁹Ar ages of >2.6, 1.8–2.0, and 1.7–1.9 Ga, respectively (Roy et al. 2007a). Given the location of the study area relative to these bedrock provinces, hornblende grains with a ⁴⁰Ar/³⁹Ar age that is >2.6 Ga (Superior Province) would indicate an eastern to southeastern provenance, since there is no evidence of east- or northeast-trending ice flow in the stratigraphic record of the western HBL (Gauthier et al. 2019). Four samples were chosen from select stratigraphic units as well as a fifth sample from a till with a well-constrained northwestern provenance near the town of Gillam (Moondance section on Fig. 1). This reference “Sundance Till” has northwest–southeast trending till fabrics with southeast-trending striations on bedrock at base of the unit and contains elevated granitoid-clast concentrations, indicating a northwestern provenance for the sediment (Nielsen et al. 1986; Roy 1998; Hodder et al. 2017). Detrital hornblende grains were handpicked from till mineral separates for ⁴⁰Ar/³⁹Ar age determination. A total of 14–16 grains were analyzed for each sample at the Ar-Ar lab at the Manitoba Isotope Research Facility. The complete ⁴⁰Ar/³⁹Ar age dataset from this study is provided in Table SB1.

The <63 µm size-fraction of till samples were separated by sieving, and then digested in a four-acid (HF:HNO₃:HClO₄:HCl) mixture. Following this near-total digestion, samples were analyzed by inductively coupled plasma mass spectrometry (ICP-MS) for 58 elements at Activation Laboratories Ltd. Exploratory data analysis of the near-total digestion dataset indicated significant variation between the 2016 and 2017 dataset (supplementary material C) that could not be explained by geological field variability; the 2017 survey increased the density of sampling but did not expand the spatial bounds of the 2016 survey. The variability between analysis years is likely due to differences in the strength of the four-acid digestion used. To perform statistical analysis of the dataset for both years together, the 2017 dataset was levelled to 2016 data. A linear regression-based leveling approach was used (as described in Grunsky 2010), correcting the 2017 data with respect to the 2016 data following an assumption about incomplete digestion. Levelling was performed using a linear regression equation derived from the comparison of every 5th percentile (e.g., 5th, 10th, 15th up to the 95th). The limited number (n = 8) of till samples collected during the 2019 field season were analyzed using a stronger lithium metaborate fusion digestion and were excluded from analysis.

Compositional data, like till-matrix geochemistry datasets, pose challenges due to their high dimensionality, hindering relationship identification and data interpretation (Grunsky 2010). To investigate whether the till-matrix geochemistry supports till-provenance interpretations, and whether it could be used to elucidate differences between tills deposited by similar ice-flow orientations, principal component analysis (PCA) was applied to the dataset. PCA addresses the high dimensionality problem by transforming the original variables into a set of orthogonal dimensions (principal components (PCs)). These components are ordered by their importance in explaining variance in the data. Focusing on the first few PCs enables concise data summarization without losing too much information (Jolliffe and Cadima 2016). PCA facilitates the detection of relationships and correlations not readily apparent in the initial dataset, which may reveal patterns of geological significance. Recently, PCA has seen increased use in analyzing sediment geochemical datasets for process discovery and exploration of specific elements or minerals associated with ore deposits (Chen et al. 2019; Bronac de Vazelhes et al. 2021), or for regional bedrock provenance analysis (McMartin et al. 2016; Lipp et al. 2021; Rice et al. 2024), which is how PCA is applied in this study. The levelled dataset was first investigated using exploratory data analysis (probability plots and histograms) and investigated for outliers (supplemental C). The dataset was then transformed using a centred log ratio transformation to avoid the effects of data closure (Grunsky 2010), i.e., variables that sum to a constant 100%. PCA was applied to this transformed dataset using ioGAS^TM software. To maintain acceptable degrees of freedom within the dataset of 109 samples, PCA was restricted to major elements (Ca, Mg, Na, Al, Fe, K, and Mn), Rb, Zr, and the sum of light rare earth elements (LREE = La, Ce, Pr, Nd, Sm, Eu). There are thus about 11 times more samples than selected variables, which should help properly condition the covariance matrix.

Fossil pollen grains have routinely been used to reconstruct paleoclimate during past nonglacial intervals in the western HBL (Netterville 1974; Nielsen et al. 1986; Dredge et al. 1990; Roy 1998). Such information can be valuable for understanding local and regional vegetation during past interstadial/interglacials and the environments (climate) inferred from them. For this study, pollen analysis was conducted on a previously undocumented organic-bearing unit. Three samples were taken from the subtill organic-bearing sediments exposed at Section 16-416 at 20–30 cm intervals, consistent with the sampling intervals at the Echoing River tributary section (Dredge et al. 1990). From each sample, a 1 cm³ subsample was used for pollen analysis. Pollen was concentrated using standard palynological techniques involving acid digestions (Faegri and Iversen 1975), supplemented with the use of sodium polytungstate as a dense floatation medium to remove mineral sediments that remained after standard pollen preparation (Zabenskie 2006; Campbell et al. 2016). Ceramic palynospheres were added to the preparation to estimate pollen concentration (Kitaba and Nakagawa 2017). Pollen samples were processed at the Paleoecology Laboratory at the University of Toronto.

In each sample, a minimum of 150 herb, arboreal, and shrub pollen grains were counted using a compound microscope at 400× magnification and identified using the pollen key for the Great Lakes region (McAndrews et al. 1973). Stratigraphic plots were made in R package “rioja” (Juggins 2015; Juggins et al. 2015). Quantitative paleoclimate reconstructions were undertaken using the modern analogue technique (Overpeck et al. 1985) while using a previously published modern-day dataset for North America (4882 sites, see Whitmore et al. 2005; Dalton et al. 2017). Statistical tests on the modern-day pollen database show strong significant relationships with total annual precipitation and mean summer temperature (June, July, and August); therefore, these were the variables targeted for paleoclimate reconstruction. Raw pollen counts and additional details about the modern analogue technique are available in Dalton et al. (2022b). To provide additional context on local soil development, we estimated organic matter content for each pollen interval using loss-on-ignition by combusting sediment at 550 °C for 4 h and measuring the mass lost (Heiri et al. 2001).

The optical age of the stratigraphically-youngest intertill sorted sediments has previously been determined for two sites in the study area (Hodder et al. 2023). This study used the same methodology to investigate a stratigraphically older nonglacial unit in the record and sample properties are provided in Table 1. Sorted sand from this unit were sampled for optical age determination by inserting an opaque plastic tube (∼20 cm long and 5 cm diameter) into the sediment face, extracting the tube, and then sealing each end to preserve water content. Samples were prepared using standard laboratory methods and 180–250 µm diameter quartz grains were isolated for dating. Equivalent dose (D_e) values were determined using multi-grain aliquots (50–100 grains per aliquot) and the single-aliquot regenerative dose method (Murray and Wintle 2000, 2003). Representative D_e values used for dating were calculated using the central age model and minimum age model (MAM) (Fig. 5; Galbraith et al. 1999). Additional details regarding the optical dating methodology can be found in Hodder et al. (2023).

A hybrid-stratigraphic approach was used that incorporates both lithostratigraphy and allostratigraphy principles to define informal stratigraphic units. The North American Stratigraphic Code (North American Commission on Stratigraphic Nomenclature 2021) recommends defining informal units when the extent and significance of units beyond the study area are still undetermined and/or when units are defined using innovative techniques. To build a stratigraphic framework of informally defined till units, a multi-parameter lithostratigraphy approach was taken that incorporates sedimentology observations (relative consolidation, colour, sedimentary structures, clast fabric, and other ice-flow indicators) and laboratory analytical data (till provenance). The approach started in the field where till beds may be recognized at a section based on facies characteristics, their upper and lower boundaries, as well as their general stratigraphic position. This allowed for preliminary facies or litho-correlations of units with similar characteristics and stratigraphic positions at a local scale. It is important to note here that this initial field-based correlation between sections incorporates ice-flow indicator data for each identified till unit. These field-based observations were verified by detailed provenance analysis (clast lithology counts and detrital hornblende ⁴⁰Ar/³⁹Ar data) to confirm the direction of ice flow that deposited the sediment. After till units were defined using a lithostratigraphic approach, an allostratigraphic approach was used to identify the relative age of intertill sorted sediment packages using the bounding surfaces of the enclosing till units. Use of an allostratigraphic approach to identify the relative age of intertill sorted sediments was necessary due to the fragmented nature of the stratigraphic record in the region, where the uppermost intertill nonglacial sediments encountered at a section may not necessarily be related to the penultimate deglaciation of the region (e.g., Dube-Loubert et al. 2013). The stratigraphy was first developed at local scales, and correlations were progressively extended across the study area to provide a regional stratigraphic framework.

Below, we discuss our stratigraphic findings at 22 sites (Fig. 6A; Hodder et al. 2024), beginning with the stratigraphically oldest recognized depositional event. The regional ice-flow data, geochronology, and spatial location of each identified stratigraphic unit are plotted in Figs. 6B–6H. Eight stratigraphic units are recognized and described below and placed in a chronostratigraphic context based on age constraints from intertill sorted sediment units.

Unit 1 occurs at seven sections (Fig. 6B) and is best exposed at Section 19-628 (Fig. 7). Unit 1 consists of a massive, matrix-supported diamicton that contains striated, bullet-shaped, and faceted clasts. The matrix is blue–grey, greenish-grey, dark greyish-brown, brown, or reddish brown in colour. The diamicton is highly consolidated and strongly jointed, with oxidation rinds developed on joint surfaces. The massive structure, clast shape, lodged in situ cobbles with striated upper surfaces, and lateral continuity indicate that Unit 1 diamicton is a subglacial traction till (Evans et al. 2006; Evans 2018). Ice-flow data from Unit 1 include lodged cobbles and spread-unimodal till fabrics that are interpreted to have formed by southwest-trending (229°–246°) ice flow (Table 2; Fig. 6B). Unit 1 till has decreased UGG clast concentrations relative to other tills (Fig. 8) and no oolitic jasper or Dubawnt clasts. The lower contact of Unit 1 was not observed at any section. The upper contact of Unit 1 is unconformable or transitional and it underlies the sorted heterogeneous sediments of Unit 2 (six occurrences) or postglacial sediments of Unit 8 (one occurrence). In the latter case, the section is located near Shamattawa within the postglacial drainage corridor (Fig. 2).

Unit 2 is recognized at eight sections (Fig. 6C) and is best exposed at Section 17-522/19-630 (Figs. 9, 10A, and 10B); this includes the previously published “Echoing River Section” (Dredge et al. 1990). Unit 2 consists of several contrasting lithofacies and is defined by its bounding surfaces (allostratigraphy). Unit 2 is defined by its upper bounding surface with Unit 3 till (one occurrence) or Unit 4 till (seven occurrences) and lower bounding surface with Unit 1 till (six occurrences). Unit 2 ranges in thickness from 0.1 to 8.5 m and consists of bedded clay, silt, sand, and gravel. Three of the occurrences (16-416, 17-522 and 17-531) are organic bearing (wood and/or peat; Fig. 9). Each section has a unique exposure of sediment lithofacies, and lateral variations across and between exposures are common. This includes the transition from organic-bearing sediment (Fig. 10A) to organic-barren sediment (Fig. 10B) or pinching out laterally into a till–till contact/unconformity (Fig. 10D). The thickness of Unit 2 changes over short distances and in many cases is the uppermost occurrence of intertill nonglacial sediments at a section (e.g., Fig. 9). Sediments correlated to Unit 2 are situated at elevations between 69–114 m asl.

The pollen record from Unit 2 was investigated at Section 16-416. Only two of the three examined samples contained sufficient pollen concentrations and unidentified grains (>5000 pollen grains/cm³ and <10% broken/unidentified) for further analysis. These two intervals contained pollen groups typical of a boreal peatland environment akin to what exists there today (Fig. 11). Notably, pollen was composed of arboreal (Picea, Pinus, Salix) and wetland (Cyperaceae and Spahgnum) components, along with grass (Poaceae) pollen, fern (Polypodiaceae) spores, and the algal aquatic indicator, Pediastrum. Loss-on-ignition analysis confirms a highly minerogenic component (organic content was 4%–6%). Quantitative paleoclimate reconstruction using the modern-analogue technique indicates deposition during a mean summer temperature of 13.4–13.9 °C, which is indistinguishable from present-day mean summer temperature of 13.4 °C (Fig. 11). Reconstructed mean annual precipitation ranged from 537 to 689 mm, which is similar or slightly higher than present-day values for the area (474 mm). Detrital wood collected from sandy gravel sediments at Section 16-416 yielded a nonfinite radiocarbon age (>56 ka ¹⁴C BP; UOC-9641; Hodder et al. 2023).

Pollen and paleoenvironmental data at section 17-522 (Figs. 9 and 10A) were previously published as the “Echoing River Section” (Dredge et al. 1990). The lowermost clay in this section was possibly deposited in a marine or glaciomarine environment. This interpretation is based on relatively elevated boron and vanadium values of the sediment and the recovery of trace foraminifera; however, these microfossils may be allochthonous (Dredge et al. 1990). Only 0.15 m of this bed was accessible when the site was revisited, and excessive colluvium prohibited further investigation. The overlying clay, silt, sand, and peat (Fig. 10A) contained freshwater gastropods and ostracods, which indicate deposition in a terrestrial or freshwater environment (Dredge et al. 1990). Similar to Section 16-416, pollen at Section 17-522 was largely composed of arboreal (Picea, Pinus) and wetland (Cyperaceae, Sphagnum) components. The pollen profile indicates an overall warming trend from an open tundra to boreal forest-type environment. The occurrence of trees in the area is corroborated by the recovery of spruce wood from the organic-rich bed (Dredge et al. 1990). The insect fauna investigated from the organic-rich bed indicates boreal conditions close to the tree line (Dredge et al. 1990). Wood at this site yielded nonfinite radiocarbon ages (>31 ¹⁴C ka BP (GSC-892); >51 ¹⁴C ka BP (GSC-4444HP); Dredge et al. 1990). At a location 160 m to the north along the same Section 17-522 exposure, Unit 2 sorted sediments consist of >1.5 m of well-sorted medium sand that is weakly horizontally bedded to cross-bedded (Fig. 10B). The optical age of the well-sorted, subhorizontally bedded medium sands were determined to be 191 ± 29 ka (MAM, 1σ (sample 112-19-630-OSL025); Table 3) or 272 ± 22 ka (CAM, 1σ (sample 112-19-630-OSL025); Table 3). This sand sample has a relatively high overdispersion value of 30 ± 5% and therefore most likely consists of grains that have been heterogeneously bleached prior to deposition. In this case, the MAM D_e would give an approximation of true depositional age of a sample if a significant number of grains had been fully bleached, or it would give an age closer to the true age than that derived using the CAM if this is not the case. For these reasons we interpret that MAM age of 191 ± 29 ka (1σ (112-19-630-OSL025)) to be closer to the depositional age of the sediments. Nonetheless, the CAM and MAM ages for this sediment bracket the MIS 7 interglacial period (243–191 ka; Lisiecki and Raymo 2005). These observations suggest that Unit 2 sediments were deposited during the MIS 7 interglacial period in marine, fluvial, lacustrine, and/or eolian environments during an overall warming trend from an open tundra to boreal forest-type environment.

Unit 3 is documented at eight sections (Fig. 6D) and is best exposed at sections 16-430 (Fig. 12) and 17-515. Unit 3 consists of a massive, matrix-supported diamicton that contains striated, bullet-shaped, and faceted clasts. The matrix is greyish brown, dark greyish brown to brown in colour. The diamicton is highly consolidated and strongly jointed with oxidation rinds developed on joint surfaces. The massive structure, clast shape, lodged in situ cobbles with striated upper surfaces and lateral continuity indicate that Unit 3 diamicton is a subglacial traction till (Evans et al. 2006; Evans 2018). Unit 3 has elevated UGG clast concentrations (eastern provenance) relative to other tills (Fig. 8). Three oolitic jasper clasts (eastern provenance) and no northern provenance Dubawnt clasts were counted (Fig. 8). The detrital hornblende ⁴⁰Ar/³⁹Ar ages are predominantly 2.6–2.7 Ga, with 10 of 15 grains falling in this range (Superior Province; Fig. 13C). Together, these parameters provide supporting evidence for an eastern provenance for sediments in this unit, which is reflected in the interpretation of ice-flow data collected. Ice-flow data collected from Unit 3 include a lodged cobble and unimodal to spread unimodal clast fabrics that are interpreted to have formed by west-northwest- to north-northwest-trending ice-flow (282°–327°; Table 2; Fig. 6D). The lower contact of Unit 3 was only observed at one section (Section 19-628; Fig. 7), where the till unconformably overlies the sorted heterogeneous sediments of Unit 2. The upper contact of Unit 3 occurs with intertill sorted sediments of Unit 5 (three occurrences), Unit 6 till (two occurrences), Unit 7 till (one occurrence), and Unit 8 postglacial sediments (two occurrences). Sediments correlated to Unit 3 are spatially restricted to the south and west of the Kaskattama highland study area (Fig. 6D). Furthermore, the till unit appears to be discontinuous laterally along the southern edge of the Kaskattama highland (Fig. SA1).

Unit 4 is recognized at 11 sections (Fig. 6E) and is best exposed at Section 16-416 (Fig. SA2). Unit 4 consists of a massive, highly consolidated, matrix-supported diamicton that contains striated, bullet-shaped, and faceted clasts (e.g., Figs. 4C, 4G, and 9). The matrix is dark greyish brown to reddish brown to light brownish grey. The diamicton is highly consolidated and strongly jointed with oxidation rinds developed on joint surface. The massive structure, high level of consolidation, clast-shape, lodged in situ cobbles and boulders with striated upper surfaces, and lateral continuity indicate that Unit 4 diamicton is a subglacial traction till (Evans et al. 2006; Evans 2018). Ice-flow data collected from Unit 4 including unimodal to spread unimodal clast fabrics that are interpreted to have formed by south- to west-southwest-trending (175°–244°) ice flow (Table 2; Fig. 6E). One section (Section 16-416) shows an up-unit switch from south- to west-trending ice flow (Table 2; Fig. 6E). Unit 4 has decreased UGG clast concentrations (eastern provenance) relative to other tills, with no oolitic jasper clasts (eastern provenance) and two Dubawnt clasts (northwestern provenance; Fig. 8). The lower contact of Unit 4 was observed at seven sections, where it is in sharp contact with sorted intertill sediments of Unit 2 (e.g., Fig. 9). The upper contact of Unit 4 is sharp or transitional with Unit 5 intertill sorted sediments (two occurrences), Unit 6 till (four occurrences), Unit 7 till (two occurrences), and Unit 8 postglacial sediments (two occurrences). Sediments correlated to Unit 4 occur across the entire study area (Fig. 6E).

Unit 5 is identified at seven sections (Fig. 6F) and is best exposed at Section 16-407 (Fig. 14A; Fig. SA3). Like Unit 2, this unit consists of highly variable lithofacies and is thus defined based on its relationship to bounding units. Unit 5 is defined by its upper bounding surface with Unit 6 till (five occurrences) or Unit 7 till (two occurrences) and lower bounding surface with Unit 3 till (two occurrences) or Unit 4 till (three occurrences). At two sites no lower bounding surface was observed due to excessive colluvium. Unit 5 ranges in thickness from 0.5 m to more than 3.8 m and consists of clay, silt, sand, and gravel and no organics were observed within the sediment except for abraded shell fragments (Fig. 14). Sand facies are often ripple cross laminated (e.g., Fig. 14A) or horizontally bedded (e.g., Fig. 14B). Each section exposes a unique sequence of sediment facies, and lateral variations in an exposure were commonly observed. This includes pinching out laterally into a till–till contact unconformity across a single exposure. The thickness of Unit 5 therefore changes over short distances. The optical age of Unit 5 sediment has been determined previously at sections 19-629 and 16-407 to be 146 ± 20 (MAM, 1σ (sample 112-19-629-024); Table 3) and 166 ± 12 ka (MAM, 1σ (sample 112-16-407-019); Table 3), respectively (Hodder et al. 2023). Sediments correlated to Unit 5 are situated at elevations between 40–157 m asl.

Unit 6 is recognized at 12 sections (Fig. 6G) and is best exposed at Section 16-407 (Fig. SA3; Hodder et al. 2023). Unit 6 consists of massive to stratified, matrix-supported diamicton that contains striated, bullet-shaped, and faceted clasts. The matrix is dark yellowish brown to light olive brown. The diamicton is over-consolidated but significantly less than the underlying Unit 1, 3, and 4 tills (where observed). The diamicton is not jointed to weakly jointed with minor oxidation rinds on joint surfaces. The massive to stratified structure, clast-shape, and lateral continuity that Unit 6 diamicton is a subglacial traction till (Evans et al. 2006; Evans 2018). Unit 6 has elevated UGG clast concentrations (eastern provenance) relative to other tills and contains seven oolitic jasper clasts (eastern provenance) and two Dubawnt clasts (northern provenance; Fig. 8). The detrital hornblende ⁴⁰Ar/³⁹Ar ages of a surface and subsurface sample from Unit 6 both have predominantly 2.6–2.7 Ga (Superior Province) distributions (Fig. 13). Together, these parameters point to an eastern provenance for the till, which is reflected in the interpretation of ice-flow data for this unit (Fig. 6G). The majority of ice-flow data collected from Unit 6 in section indicates deposition by west-northwest- to northnorthwest-trending ice flow as interpreted from spread unimodal to unimodal till fabrics (287°–333°; Table 2; Fig. 6G). One spread unimodal till fabric in Unit 6 is oriented perpendicular to this orientation (201°) and is interpreted as a transverse fabric, i.e., perpendicular to ice flow, since the qualitative (sedimentology) and quantitative (clast count) properties of this till are similar to Unit 6 till at adjacent sections nearby. The streamlined landforms of the Kaskattama flowset, which trend toward the 290° on the east side of the Kaskattama highland and 305° at the northwest edge of the Kaskattama highland (Figs. 2 and 6G) are interpreted to be the surficial geomorphic expression of Unit 6. This sediment–landform relationship is evident at some sections where Unit 6 extends to the surface within the footprint of the Kaskattama flowset and is not overlain by postglacial sediments (e.g., Section 16-413). Furthermore, till collected from streamlined landforms of the Kaskattama flowset is qualitatively similar to that of Unit 6 in section (e.g., Fig. 4A) and has similar till UGG clast content (Fig. 8).

The lower contact of Unit 6 was observed at 11 sections, and it separates Unit 6 from underlying Unit 3 till (two occurrences), Unit 4 till (four occurrences), and intertill sediments of Unit 5 (five occurrences). At Section 16-407 (Fig. SA3), there is a lower clay-rich facies of the till with rip-up clasts of mud, suggesting that ice may have advanced over fine-grained sediments (Hodder et al. 2023). Similarly, underlying Unit 6 at Section 17-519 is bedded sand, silt, and clay (Fig. 14D; Fig. SA4). The lower 30–40 cm of the diamicton has the same colour as the underlying silt implying incorporation of the underlying silt into the matrix (Fig. 14D), supporting the interpretation that Unit 6 till formed during an advance over fine-grained sediments. Unit 6 till is at the surface (five occurrences) or overlain by postglacial sediments of Unit 8 (seven occurrences). Unit 6 till was encountered on the Kaskattama highland, south of Kaskattama highland underlying postglacial sediments and at one section at the junction of the Gods and Hayes rivers (Fig. 6G).

Unit 7 is interpreted at five sections (Fig. 6H) and is best exposed at Section 17-542 (Fig. SA5). Unit 7 consists of massive to stratified, matrix-supported diamicton that contains striated, bullet-shaped, and faceted clasts. The matrix is brown to dark yellowish brown to olive brown. The diamicton is over-consolidated but significantly less consolidated relative to the underlying Unit 1, 3, and 4 tills (where observed) and in places has a boulder pavement at its lower contact (e.g., Fig. 4C). The massive structure, clast-shape, and lateral continuity indicate that Unit 7 diamicton is a subglacial traction till (Evans et al. 2006; Evans 2018). Ice-flow data collected from Unit 7 include lodged boulders and unimodal to spread unimodal till fabrics that are interpreted to have formed by south- to west-southwest-trending ice flow (176°–240°; Table 2; Fig. 6H). The streamlined landforms of the Hayes flowset are oriented 204°–207° in the study area (Fig. 2) and are interpreted to be the surficial geomorphology expression of Unit 7 based on similarities between clast fabric direction and streamlined landforms, stratigraphic position, and qualitative similarities between the surficial till and till described in section. Sections 16-416 and 17-529 have strong south-trending fabrics (176° and 178°; Table 2), and Section 16-416 is overlain by the Hayes streamlined-landform flowset (Fig. 6H). As such, formation of the south-trending clast fabric occurred prior to the late-glacial south-southwest-trending ice flow that formed flowset. Unit 7 has average UGG clast concentrations (eastern provenance) relative to other tills and contains one oolitic jasper clasts (eastern provenance) and five Dubawnt clasts (northern provenance; Fig. 8). The five Dubawnt clasts counted amount to 55% (n = 5/9) of the total count in the dataset, yet Unit 7 samples represent only 15% of the sample population. The detrital hornblende ⁴⁰Ar/³⁹Ar age distribution from one sample is predominantly 1.7–2.0 Ga (Fig. 13; WC or THO, Fig. 3). This ⁴⁰Ar/³⁹Ar age distribution is similar to the Sundance Till situated to the west of the study area near the town of Gillam (Fig. 1), which has a well-documented northwestern provenance (Fig. 13). Together, these provenance parameters indicate a northwestern provenance for Unit 7. This contrasts with the orientation of the surficial flowset indicative of southwest ice flow and till fabrics indicative of south and southwest ice flows. Considering these ice-flow directions, the northwest provenance indicators in Unit 7 could be inherited from older tills deposited by southeast-trending ice flow or could indicate long-distance transport from mainland Nunavut by the Hayes Lobe. The lower contact of Unit 7 was observed at all five sections, with Unit 3 till (one occurrence), Unit 4 till (two occurrences), and intertill sediments of Unit 5 (one occurrence). The upper contact of Unit 7 is overlain by postglacial sediments of Unit 8 (three occurrences), or the unit ends at the surface of the section (one occurrence) and could not be accessed for observation at one section. Unit 7 till is spatially restricted to the Hayes flowset or in areas overlain by Unit 8 postglacial sediments in close proximity to the flowset.

Unit 8 consists of massive to bedded clay, silt, sand, and gravel that was deposited during the Holocene Epoch (MIS 1) following retreat of LIS from northeastern Manitoba. These sediments were deposited in glaciolacustrine, marine, or fluvial environments and have collectively been grouped into one stratigraphic unit that is representative of “postglacial sediments”. A detailed investigation of the postglacial history of this region is beyond the scope of this manuscript; however, some aspects of the postglacial history of this region are discussed in Dredge (1983) and Gauthier et al. (2020).

PC 1, 2, and 3 account for 40%, 19%, and 15% of the observed variance within the dataset, respectively (Fig. 15A). Exploratory examination of PC biplots indicated that stratigraphic units, identified independent of the matrix geochemical data, were best differentiated using PC1 and PC3 (Fig. 15B) and the biplot of PC1 and PC2 is provided in Fig. SD1 for comparison. Samples with positive PC1 values are representative of a relative Shield-rich provenance signature (Al–Fe–Rb; e.g., Fig. 15C). Samples with negative PC1 values and positive PC3 values are representative of a relative Paleozoic-rich provenance signature (Ca–Mg–Mn; e.g., Fig. 15D). Samples with negative PC3 values are representative of a relative UGG-rich provenance signature (Na–Zr; e.g., Figs. 15C and 15E).

This study incorporated a hybrid stratigraphic approach that identified till units using principles of lithostratigraphy. We then used principles of allostratigraphy and examined the bounding surfaces of identified till units with intertill sorted sediments. This was a necessary approach since the stratigraphy is dominated by glacial sediments (till) and intertill sorted sediments that are heterogeneous and variably preserved (or deposited) in the stratigraphic record. This approach assembled a stratigraphic framework for the region consisting of eight units (Fig. 16). The units are Unit 1, a till deposited by southwest-trending iceflow; Unit 2, nonglacial sediments of likely MIS 7 age; Unit 3, a till deposited by northwest-trending ice flow; Unit 4, a till deposited by south- to west-trending ice flow; Unit 5 nonglacial sediments of early MIS 5 or late MIS 6 age; Unit 6, a till deposited by northwest-trending ice flow; Unit 7, a till deposited by south- to southwest-trending ice flow; and Unit 8, sorted sediment of MIS 1 age.

Deciphering the fragmented record from the Kaskattama highland, especially identifying till units, is challenging because the stratigraphic framework relies on 22 sections and is not complete at any one location in the study area (Fig. 17A), which is a common theme in regions of thick till-dominated sequences (e.g., Möller et al. 2019). This highly fragmented record highlights either patchy deposition of till, patchy erosion of till, or likely a combination of both processes. This work shows that till units in the region can be indistinguishable based on certain properties (e.g., texture, colour, and compaction) yet have contrasting till fabrics and unique provenance indicators, which provide compelling evidence they were transported and deposited by ice flowing in different directions (Fig. 4D vs. 4E and 4F vs. 4G). “Similar tills”, such as Unit 3 and 4, can be in the same relative-stratigraphic position (underlying nonglacial Unit 5) and spatially close (Fig. 17A; Fig. SA1), but be deposited by ice flowing in near-perpendicular directions. To differentiate these similar till units, our approach relied upon the till clast-fabrics coupled with till provenance data to determine the ice-flow direction at time of till deposition, and the dominant bedrock source regions of the sediment particles. This is an important distinction, because previous stratigraphic work in the western HBL did not define tills by ice-flow direction, but instead the ice-flow data were considered an attribute of the lithostratigraphic units. In some cases, that led to identified till units containing mismatched ice-flow indicators (e.g., Long Spruce till of Nielsen et al. 1986). To accurately construct a Quaternary stratigraphic framework in similar areas, the multi-parameter characterization of till units, using parameters such as clast fabric, colour, relative degree of compaction, clast lithology, and till-matrix geochemistry, is necessary and an emphasis needs to be placed on collecting extensive ice-flow indicator datasets.

Understanding the till stratigraphy is also a necessary step to establish the relative age of intertill sorted sediment units using an allostratigraphic approach. If we examine the current stratigraphic architecture of the study area, Holocene-aged postglacial sediments (Unit 8) overlie till correlated to five separate units deposited during at least three separate glaciations (Fig. 17B). Age constraints on the intertill sediments (Unit 2 and 5) suggest that these tills were deposited during MIS 2–5, 6 and 8, or older (Fig. 16). Previously, Unit 2 sediments at section 17-519 were interpreted to be MIS 5 in age (Dredge and McMartin 2011) based on the paleobotanical reconstructions that indicate deposition during an interglacial climate and because the interglacial bed was the first subtill occurrence of organic-bearing sediments. This new stratigraphic framework and optical age estimation suggests that these sediments were likely deposited during MIS 7 (191 ± 29 ka, MAM, 1σ (112-19-630-OSL025)). This means that, as in other HBL stratigraphic studies, the uppermost sorted sediments underlying glacial sediments (till) may not have been deposited during the penultimate ice-free period (e.g., Dube-Loubert et al. 2013; Gauthier et al. submitted), and therefore cannot be used to “anchor” a stratigraphic framework or used as “marker” beds for correlating between stratigraphic observations.

It must be noted that the optical age estimations from Unit 2 and 5 slightly overlap at 1σ error (Table 3) and this reflects the difficulties in dating nonglacial deposits > 100 ka. Nonetheless, this study has documented two distinct intertill nonglacial sorted sediment units separated by till(s) indicating a glacial event(s). The simplest interpretation based on our new Quaternary stratigraphic framework, existing age constraints, and comparison to the marine oxygen isotope record is that Unit 2 was deposited during MIS 7 and Unit 5 was deposited during MIS 5–6. The age of intertill sorted sediment units presented in the stratigraphic framework herein can readily be tested as geochronology methods improve since unit correlations did not rely on the estimated ages determined.

Using ice-flow data, till provenance, and stratigraphic relationships we show that the enigmatic Kaskattama streamlined-landform flowset is a part of stratigraphic Unit 6 that extends off the Kaskattama highland, and the flowset was formed by ice flowing toward the northwest (Fig. 6G). At several sections investigated (e.g., 16–416 and 17–519), the stratigraphy indicates that this ice-flow event advanced over a fine-grained substrate and incorporated these sediments into the overlying till, including mud rip-up clasts. This ice-flow phase is interpreted as an advance-phase event because it requires an ice sheet configuration that is consistent with other reconstructions that put a Q–L ice divide east of James Bay in the early build-up phase and MIS 4 (e.g., Veillette et al. 1999). The timing of this ice-flow event is constrained by the optical age estimations of the underlying intertill sorted sediments (Unit 5) of 146 ± 20 ka (MAM, 1σ (112-19-629-024)) and 166 ± 12 ka (MAM, 1σ (112-16-407-019)). A third optical age of 150 ± 22 ka (MAM, 1σ (112-19-605-004)) was determined from correlative subtill sediments at a site near the Manitoba–Ontario border (Hodder et al. 2023; Black Duck River site on Fig. 1). Here, 1.4 m of till with a west-trending (273°) till fabric is underlain by massive silt and massive to horizontally-bedded fine-sand to silty fine-sand that contains disseminated organics (Hodder et al. 2023). These sediments were correlated to the Black Duck River Section of Dalton et al. (2016), where the pollen signature indicates deposition during climatic conditions similar to the current interglacial period (Dalton et al. 2022b). As such, Unit 6 till advanced over sorted sediments (Unit 5) interpreted to have been deposited during the MIS 5 interglacial period or MIS 6 deglaciation, and records an initial advance of ice emanating from the Q–L dome into the western HBL at the start of the last glacial cycle (Fig. 16). This glacial advance likely occurred during MIS 4 or late MIS 5, based on global sea level considerations and modelling of the LIS (Stokes et al. 2012; Gowan et al. 2021) and other regional field-based studies (e.g., Veillette et al. 1999).

The presence of till and streamlined landforms at the surface, interpreted to have been deposited at the start of the most recent glaciation, implies that this region experienced extensive landscape preservation throughout the vast majority of the last glaciation, possibly under cold-based conditions, a phenomena observed in other inter-ice stream localities (Kleman et al. 1994; Hättestrand and Stroeven 2002; Kleman and Glasser 2007; Ross et al. 2009, 2011). The Kaskattama highland rises ∼130 m above the flat-lying HBL and the highland is flanked by deglacial ice streams to the east (Winisk ice stream) and west (Hayes Lobe). The interpretation of the Kaskattama flowset as an advance-phase till supports the concept that the highland was likely a large inter-ice stream sticky spot during deglacial ice streaming in the region (Stokes et al. 2007; Gauthier et al. 2019). This work now conclusively shows that these fragmented landscape patches, or glacial terrain zones (c.f. Trommelen et al. 2012), can be of different ages and represent spatiotemporal variations within the subglacial bed mosaic (c.f. Gauthier et al. 2019) and that this needs to be considered when reconstructing the depositional record over multiple glacial cycles.

This study shows that two episodes of northwest-trending ice flow advanced over the study area at a similar trajectory during different glacial cycles (Unit 3 and Unit 6; Fig. 16). This new stratigraphic framework places the oldest northwest-trending ice-flow event at the start of MIS 6. Importantly, this framework is supported by regional-scale correlations in the western HBL (Gauthier et al. 2019) and recent stratigraphic observations in the Churchill River area (Hodder and Gauthier 2023). Within the central HBL, a relatively old till deposited by northwest-trending ice flow referred to as the Rocksand Till is recognized (Thorleifson et al. 1992). The timing of deposition of the Rocksand Till was interpreted to be either at the start of last glacial cycle (late MIS 5) or MIS 6 glaciation and this interpretation was dependent on the accuracy of available thermoluminescence (TL) ages of overlying sediment at the time (Thorleifson et al. 1992). The TL ages incorporated by Thorleifson et al. (1992) from Manitoba are now interpreted as age underestimations and it is likely that these sediments were deposited during MIS 5 instead of MIS 3 based on new geochronology constraints (Hodder et al. 2023), therefore placing the Rocksand Till in MIS 6. This alternative framework presented in Thorleifson et al. (1992) would also re-assign the Missinaibi Formation, including the Bell Sea sediments, to the MIS 7 interglacial period. This reinterpretation has significant implications for the relative age framework proposed using amino acid racemization data for the HBL (Andrews et al. 1983), since this framework assumed the Bell Sea Sediments were deposited during MIS 5e. Furthermore, the Rocksand Till is tentatively correlated to Unit 3 in the Kaskattama highland area based on stratigraphic position, clast fabric orientations, and UGG clast content, which extends this northwesterly ice-flow event into the western HBL. Therefore, we suggest that this ice-flow event should no longer be used to constrain the growth of the LIS during the last glacial cycle (e.g., Kleman et al. 2010; Dalton et al. 2022a), but instead is an important regional-scale ice-flow event to be considered in reconstructions of the MIS 6 glaciation.

Despite the similar ice-flow directions for Unit 3 and Unit 6, the provenance signature of the tills deposited by these ice-flow events is different and best exemplified by their till-matrix geochemistry (Fig. 15B). Unit 6 till has a relatively high granitoid provenance signature compared to Unit 3, which is more calcareous. The nearest igneous and metamorphic rocks to the east/southeast are part of the Superior Province, which implies a higher proportion of more distally derived detritus within Unit 6 till. Interpreting the cause of the different provenance signature between these ice-flow events is difficult since the principles of overprinting and inheritance (e.g., Trommelen et al. 2013) need to be considered, as well as the overall evolution of sediment transport and deposition in the HBL. For example, one interpretation is that the Q–L dome was situated further to the east during deposition of Unit 6 relative to Unit 3, which would have led to increased erosion and transportation of detritus derived from the Superior Province into the western HBL. Alternatively, the Q–L dome could have been situated in the same position during these ice-flow events, but Unit 3 has a more local bedrock signature, possibly because it was deposited earlier in time when there was presumably more bedrock exposed or because it eroded a more Paleozoic-rich (local signature) till that was previously deposited. Relating these far-field observations to evolution of the Q–L dome is difficult without direct comparison to reconstructions from the core region of the Q–L dome, which are generally restricted to the erosional record and identified ice-flow events that are largely attributed to the last glacial cycle (e.g., Veillette et al. 1999).

This study constructed a stratigraphic framework for the Kaskattama highland region of the western HBL that consists of five glacial (till) units spanning at least three glaciations and three nonglacial (organic-bearing sorted sediments) units that were deposited during separate interglacials and supported by radiocarbon and optical age constraints. Importantly, we acknowledge that the stratigraphic record is highly fragmented, which is similarly expressed in the fragmented surficial streamlined landform record. To produce a stratigraphic framework in this geological setting, a hybrid stratigraphic approach that incorporated principles of lithostratigraphy and allostratigraphy with an emphasis on supporting ice-flow datasets was incorporated. To determine the flow direction of ice (northwest vs. southeast) that deposited till units, an extensive characterization of the sediment provenance was undertaken that included till-clast lithology counts, matrix geochemistry, and detrital hornblende ⁴⁰Ar/³⁹Ar ages to provide a more robust bedrock provenance interpretation. The application of optical dating and paleobotanical analyses to interglacial units has provided important age and climatic brackets for the timing of glacial events. Nevertheless, more work needs to be done to assess the utility of optical dating of quartz from HBL (Hodder et al. 2023; Gauthier et al. 2024). This is especially so for samples, like the one discussed here, where calculated ages are greater than 150 ka. The magnitude of the optical ages requires extensive machine time, and therefore limits the number of ages that can be provided for any one unit. Nevertheless, dating of more samples at key units should be done to better understand both variations in luminescence signals and calculated ages, possibly alongside other geochronology methods.

The implications of this stratigraphic record provide important constraints to the growth of ice sheets and glacial dynamics within the interior region of North America. At the start of the last glaciation, northwest-trending ice advanced into the western HBL and is currently exposed at surface (Kaskattama flowset). During late glacial streaming, this “patch” of the subglacial bed deposited during the advance of the ice sheet was preserved as a large “sticky spot” flanked laterally on both sides by ice streams. The existence of this patch requires extensive landscape preservation for the majority of the glaciation, implying persistent cold-based conditions. Interestingly, this study showed that there is an older northwest-trending ice-flow event in the stratigraphic record that also advanced over sorted sediments. This till unit is tentatively correlated to the Rocksand Till documented in the central HBL and recent geochronology studies now support this as a likely early MIS 6 ice-flow event. This indicates that the growth of the ice sheet during the start of the last two glaciations was likely similar with an accelerated growth of the Q–L dome early in the glaciation relative to the Keewatin dome.

Christian Böhm is thanked for fieldwork assistance during the 2017 field season. Maria Schaarschmidt, Vanessa Brewer, and Nicola Ferguson are thanked for their preparation, measurement, and analysis of optical dating samples at the University of the Fraser Valley. Elijah Deng is thanked for the preparation of samples for palynology at the University of Toronto. This article greatly benefited from thoughtful reviews by Rod Smith and David Evans.

A freely available supporting publication to this manuscript is published as Manitoba Geological Survey Open File OF2024-1 (Hodder et al. 2024), which contains the stratigraphic columns for each site investigated as well as the analytical datasets from this study in their entirety. Any additional data are available by request.

Conceptualization: TJH, MSG, SEK, OBL

Formal analysis: TJH, SEK, OBL, ASD, SAF

Funding acquisition: MR

Investigation: TJH, SEK, OBL, ASD, SAF

Methodology: TJH, MSG

Supervision: MR, OBL

Visualization: TJH

Writing – original draft: TJH, MSG, SEK, ASD

Writing – review & editing: TJH, MSG, MR, SEK, OBL, ASD, SAF

This research was funded by the Manitoba Geological Survey and Natural Resources Canada’s Geo-mapping for Energy and Minerals program (Geomapping for Energy and Minerals (GEM-2) and GEM-GeoNorth), as administered by the Geological Survey of Canada. Specifically, GEM-2 funding supported Dr. Samuel Kelley while he was a postdoctoral fellow at the University of Waterloo and GEM-GeoNorth funding supported the analysis of one optical age presented in this manuscript. NSERC Discovery Grant and Research Tools and Instrument Grant funding to OBL and a NSERC Discovery Grant to SAF.

Supplementary data are available with the article at https://doi.org/10.1139/cjes-2024-0018.