The Empress Group in Alberta, Canada Open Access

Gravel and sand deposits that overlie bedrock and underlie till are the terrestrial sedimentary record of an interval when the Canadian Interior Plains (CIP; Bostock 1970) were dominated by cycles of fluvial incision, planation, and deposition (Alden 1932; Howard 1960; Vonhof 1969; Leckie 2006). These deposits, typically referred to as basal gravel, occur across all terrains in the CIP, from the highest uplands (e.g., Cypress Hills, Wood Mountain, and Hand Hills) to the lowest buried valleys (Fig. 1). Despite the widespread distribution of basal gravel deposits and their importance as aquifers (Osborn and du Toit 1991; Barker et al. 2011; Hartman et al. 2022), most are buried and remain incompletely mapped. Consequently, except for the Saskatchewan Gravel Formation, which occupies buried valleys in southern Alberta and Saskatchewan (McConnell 1885; Stalker 1968; Whitaker and Christiansen 1972; Jackson et al. 2011), formal formation status has currently been given only to well-mapped upland units (e.g., Cypress Hills, Wood Mountain, and Hand Hills formations). It has become a common practice that elsewhere, all preglacial fluvial sediments occupying integrated buried valley networks and occurring between bedrock and the earliest till be assigned to the Empress Group (Whitaker and Christiansen 1972), apart from a limited number of areas in central and northeastern Alberta, where similar deposits are classified as the Empress Formation (Jackson et al. 2011).

Formal stratigraphic definition of upland basal gravel deposits at the formation rank has supported a comprehensive geologic history of much of the fluvial exhumation of the CIP (summarized by Leckie 2006). However, incomplete mapping and a lack of formal stratigraphic definition of deposits in lower physiographic settings have obscured the geologic history of these units. In particular, development of the buried valley network and the stratigraphic relationship between preglacial basal gravel units and continental tills is not well described. Despite understanding that buried valley basal gravel and conformably overlying proglacial lacustrine deposits in Alberta must have been emplaced between successively more extensive continental glaciations, most workers simply describe the units as time-transgressive (Jackson et al. 2011; Cummings et al. 2012) or identify Empress Group units in a range of stratigraphic positions with respect to continental tills (e.g., Evans and Campbell 1995).

In this paper, we present the results of new three-dimensional modelling which portrays the bedrock topography of Alberta and the distribution of associated basal gravel deposits and their relative positions on the bedrock surface. We examine the results of these models within the context of two regions where detailed lithologic description and identification of bounding unconformities have previously been employed to define basal gravel units at formation-equivalent stratigraphic resolution. The two areas, the Peace River region and the South Athabasca Oil Sands (SAOS) area, are located at the ends of the largest contiguous buried valley system in Alberta, which includes the Shaftesbury, Muskwa, and Wiau buried valleys (SMW; Fig. 1); thus, their stratigraphies underpin an improved understanding of buried valley evolution and provide new insights into the Quaternary history of these parts of the CIP. This work highlights the step-form physiography of the bedrock surface and the geometry of associated basal gravel units that resulted from sequential fluvial incision, lateral planation, and deposition. It also recognizes the importance of bounding unconformities in the definition of basal gravel units. With these provincial and regional scale results, we (1) discuss the geometry and genesis of basal gravel deposits, (2) synthesize the geologic history of these deposits relative to the overlying glaciogenic units with which they are chronostratigraphically intercalated, (3) propose new formal formation-rank definitions of two basal gravel units in the Peace River region, and update the current formal definitions of basal gravel units in the SAOS area, and (4) propose updates to the formal definition of the Empress Group that will improve its clarity and utility in Alberta. To avoid confusion, the long-established names of basal gravel units in both the Peace River region and SAOS area are utilized prior to the discussion where new or updated formal definitions, and consequently new names, are introduced. Formal definitions of basal gravel formations following the North American Stratigraphic Code (NACSN 2021) are presented in the supplementary information.

Fluvial exhumation of the CIP following the Early Miocene has removed up to 2 km of sediment (Nurkowski 1984; Leckie 2006). The sedimentary remnants of these ancestral fluvial systems now mantle isolated uplands throughout the CIP, which record their northeast drainage, down the regional slope of the CIP to now-beheaded deltas off the Labrador coast (Cummings et al. 2012; Corradino et al. 2022).

Cycles of fluvial incision and the deposition of basal gravel were initially attributed to the evolution of four peneplains across the CIP, represented by two upland surfaces, the modern plains surface, and the floors of buried valleys (Calhoun 1906; Collier and Thom 1918; Alden 1932; Warren 1939; Horberg 1954; Howard 1960; Edwards and Scafe 1996; Hartman 2015). According to the peneplanation model, which postulates that plains are eroded mainly by downwasting of slopes to base level (Davis 1899, 1902), each surface would be associated with a discrete fluvial deposit that is stratigraphically equivalent across that surface and therefore correlative across widely separated areas. However, the recognition that basal gravel deposits are widespread and occur across flights of broad, stepped, bedrock surfaces at more than four relative heights (Vonhof 1969; du Toit 1988) led Osborn and du Toit (1991) to propose that lateral fluvial planation (cf. Crickmay 1933, 1960, 1974) better explains the observations. Lateral fluvial planation involves the lateral migration of rivers by undercutting bedrock. Rates of incision must be small relative to rates of migration. However, the step-form physiography of gravel-mantled bedrock surfaces indicates that periods of lateral migration are punctuated by intervals of more rapid incision. Osborn and du Toit (1991) provide examples of lateral fluvial planation at scales between tens of metres to hundreds of kilometres and note that the migration of ancestral fluvial systems across the CIP documented by Vonhof (1969) is explained by the process. Leckie (2006) documents over 700 km of lateral migration of ancestral Missouri River between the Cypress Hills and southern Manitoba (Fig. 1), which he argues is the primary architect of the modern CIP landscape.

Interpretation of bedrock physiography and the distribution of basal gravel as products of lateral fluvial planation is significant since it provides a mechanism for the incision of progressively lower valleys, along which successively younger basal gravel units would be deposited and subsequently buried by glaciogenic sediments spanning several glaciations. These cycles of fluvial incision and lateral planation were likely watercourse-specific; therefore, stratigraphic correlations of basal gravel deposits occupying widely spaced buried valleys based only on similar relative physiographic positions would not be possible.

Early reconstructions of continental ice sheet extent proposed that the first (pre-Illinoian) glaciation was the most extensive across western parts of the CIP (Stalker 1963, 1969, 1983; Stalker and Foster 1976; Stalker and Harrison 1977; Stalker and Wyder 1983). However, Middle Wisconsinan radiocarbon dates from sub-till gravel deposits and overlying proglacial sediment infilling buried valleys in west-central Alberta (Westgate et al. 1971, 1972; Catto 1984; Liverman et al. 1989; Young et al. 1994; Catto et al. 1996), and cosmogenic dates of the Foothills Erratics Train (Jackson et al. 1997; Fig. 1) indicate that the only time continental ice extended to west-central, central, and southwest Alberta occurred during the Late Wisconsinan, coincident with the all-time maximum of the Laurentide Ice Sheet (Jackson et al. 1999; Fig. 1). Jackson et al. (2011) noted that Illinoian and Early Wisconsinan ice sheets only extended into eastern Alberta, supporting a model of progressive westward expansion of continental ice sheets through the Pleistocene (Barendregt and Irving 1998; Batchelor et al. 2019; Dalton et al. 2022). Therefore, despite their contiguous appearance, the major buried valleys spanning Alberta (Fig. 1) evolved sequentially, since gravel emplacement along their floors was followed by the deposition of up to seven continental tills in Saskatchewan (Barendregt et al. 2012), five in northeastern Alberta (Andriashek and Barendregt 2017), two or three in southeastern Alberta (Proudfoot 1985; Evans and Campbell 1995; Jackson et al. 2008), but only one in west-central and southwestern Alberta (Young et al. 1994; Catto et al. 1996; Jackson et al. 2008). Thus, basal gravel has been deposited within paleovalleys across Alberta between at least the Early Pleistocene (MIS 21) and Middle Wisconsinan (MIS 3;Fig. 2).

Similarly, the distribution of Cordilleran-derived tills east of the Rocky Mountain front between northwest Alberta and Montana records successive advances of montane glaciers onto the CIP between the Late Pliocene and Late Wisconsinan (Moran 1986; Cioppa et al. 1995; Karlstrom 2000; Bobrowsky and Rutter 2007; Atkinson et al. 2016; Utting et al. 2016; Hartman et al. 2018; Figs. 1 and 2). These advances contributed outwash to the western reaches of successive basal gravel units, as reported along buried valleys near Lethbridge and along the Peace River (Wagner 1966; Jackson et al. 2008; Hartman et al. 2018). However, glacially derived carbonate clast characteristics of montane outwash are rare in older basal gravel deposits, presumably due to abrasion or dissolution (Jackson et al. 2011).

Empress Group sediments are defined as the lowermost deposits of integrated buried valley networks, comprising coarse quartzitic gravel of fluvial origin conformably overlain by laminated to bedded sand, silt, and clay (Whitaker and Christiansen 1972). Upward fining is the regional signature of drainage impoundment by the advance of continental ice sheets and is accompanied in some areas by the first occurrence of Canadian Shield clasts (pink to red granite and granitoid gneiss) as dropstones and diamictic pods below the base of overlying till (Proudfoot 1985; Evans and Campbell 1995). Basal gravel units that mantle plains adjacent to buried valleys typically do not fine upward were not emplaced immediately prior to glaciation and are therefore excluded from the Empress Group (Evans and Campbell 1995). However, as outlined by Whitaker and Christiansen (1972) and Evans and Campbell (1995) the lithostratigraphic definition may group contiguous, lithologically inseparable basal gravel bodies that can include pre- and post-glacial deposits.

The initial development of the stratigraphic framework for the Empress Group was based on the premise that once fluvial incision had established the valley floor peneplain, it was contemporaneously infilled by preglacial gravel, and then buried beneath the first and most extensive continental till sheet that extended across the CIP. However, modern understanding of the fluvial and glacial histories of the CIP indicates a far more complex stratigraphy in which the first occurrence of glaciogenic sediments at the top of the Empress Group is spatially and temporally transgressive. Basal gravel units must therefore be similarly diachronous. While the assignment of preglacial basal gravel to the Empress Group provides a useful local stratigraphic marker to establish the first arrival of continental ice into an area, such assignments become increasingly unworkable at the provincial scale.

The spatial distribution of basal gravel across Alberta was estimated using a three-dimensional probabilistic machine learning approach. Lithological descriptions from 335 970 boreholes including not only predominantly water wells, but also geological test holes, and 6613 field sites or stratigraphic sections (Fig. 3, inset A) were coded into gravel and/or sand, and clay classes based on pattern matching of character combinations in the interval descriptions. The location (x, y) of the boreholes/sites and the mid-point depths (z) of each interval were interpolated as a categorical/classification model in three dimensions, using the Extreme Gradient Boosting (XGBoost) machine learning algorithm (Chen and Guestrin 2016). The algorithm was optimized using a randomized search approach by tuning the learning rate (eta), maximum depth of each tree (max_depth), the proportion of predictors that are randomly selected to be available for each node split (colsample_by_tree), and the size of the bootstrap sample (sample_size) using 25 randomly selected parameter combinations.

To incorporate knowledge of depositional processes and improve the predictive performance of the model, point data locations were augmented by information derived from terrain physiography (elevation, slope, topographic roughness, and relative heights between valleys and/or ridges), geologic setting (sediment thickness), and spectral properties of surface materials using MODIS (moderate resolution imaging spectroradiometer) surface reflectance satellite data (Vermote 2015). Three-dimensional proximity measures to the thalwegs of mapped buried valleys, using the depth values from the bedrock surface model, were also used as a basis for predicting lithological changes that are potentially associated with deposition within the paleovalley network. The model predictions depict the probability of gravel and sand occurrence in the sub-surface in three dimensions.

The predictive performance of the model was also estimated using a train-test split with 20% of the boreholes/sites being used as a holdout set, with the model being trained on the remaining 80%. Classification performance was evaluated using standard metrics, with the area under the curve of the receiver operating characteristic (AUC-ROC) being used as a measure of overall performance. Sensitivity (sens) was also used to measure the model’s ability to predict positive (gravel/sand) results out of the total number of positive cases, and specificity (spec) was used to measure the proportion of negatives (clay) that are correctly classified as negative.

We review basal gravel stratigraphy in two regions on either end of the largest buried valley system in Alberta (SMW buried valley system; Fig. 1) where the lithology of basal gravel units has been previously described, and their bounding unconformities with underlying bedrock straths and overlying glaciogenic deposits are mapped at formation-equivalent resolution. These are the Peace River region between northeastern British Columbia and northwestern Alberta, and the SAOS area of northeastern Alberta (Fig. 1). Since discriminating between Empress Group gravel and younger alluvium can be lithologically challenging (Vonhof 1969; Whitaker and Christiansen 1972; Evans and Campbell 1995), our approach combines allostratigraphy (e.g., the physiography of the underlying bedrock unconformity, and age and origin of overlying glaciogenic sediments) with lithostratigraphy, as advocated by Räsänen et al. (2009) to describe thin, spatially, and lithologically variable units in glaciated terrains.

The trained model yielded moderately predictive performances based on the holdout set (AUC-ROC = 0.76). Model performance is better in areas of higher data density (Fig. 3, inset A) or in regions that contain a simpler stratigraphy consisting of one or two dominant units of coarse- or fine-grained deposits. Areas containing thick sediment sequences and complex sub-surface stratigraphy have moderate predictive accuracies. A quantitative evaluation of model performance is provided in the supplementary information.

Figure 3 (inset B) shows the probability of gravel and sand occurrence at 2 m above bedrock. Within ∼300 km of the mountain front, where bedrock elevation is ≥600 m a.s.l. (Fig. 3, inset C), several areas of extensive basal gravel are recognized, including (1) isolated uplands ∼200 to 600 m above the surrounding plains, (2) stepped benchlands adjacent to major rivers near the mountain front, (3) buried valleys (including valley floors and strath terraces) incised 50–150 m below the surrounding plains, and (4) plains adjacent to buried valleys. Further east, predictive modelling indicates basal gravel occupies buried valley (≤220 m below the plains), plains, and upland settings (≤600 m above the plains). In northern Alberta, data density is relatively low (Fig. 3, inset A); however, comprehensive geologic investigations in northeastern Alberta confirm the presence of basal gravel along the floors and strath terraces of buried valleys (Andriashek 2003; Utting 2021; Section 4.3).

The distribution of basal gravel across the step-form physiography of the bedrock surface is apparent in Figs. 3 and 4, which enables the relative physiographic position of discrete deposits to be established. As each strath represents an unconformity separating its associated basal gravel from units that mantle higher straths, predictive modelling of basal gravel and bedrock topography facilitates incorporation of allostratigraphic elements in stratigraphic definitions and thereby enhances the stratigraphic resolution of basal gravel units.

Four basal gravel units have been recognized in the Peace River region from northeastern British Columbia to the Vermilion lowland between the flanks of the Clear Hills and Halverson Ridge and the Buffalo Head Hills (Fig. 1). These include a gravel cap on the Clear Hills and Halverson Ridge (690 m above the Peace River), and the informally named Grimshaw, Old Fort, and Shaftesbury gravel units that mantle bedrock straths along the north side of the lowland at average heights of 320, 220, and 50 m above the Peace River, respectively (Figs. 5 and 6; Green and Mellon 1962; Mathews 1978; Hartman and Clague 2008; Atkinson and Paulen 2010; Morgan et al. 2012; Slomka and Hartman 2019). The average thicknesses of the Grimshaw, Old Fort, and Shaftesbury units are 30, 8.5, and 8 m, respectively (Slomka and Hartman 2019), with gravel on the Clear Hills and Halverson Ridge having an observed thickness of at least 8 m.

All basal gravel and sand units share a broadly similar quartzite and chert-dominated composition (Green and Mellon 1962; Edwards and Scafe 1996). However, the Grimshaw, Old Fort, and Shaftesbury gravel units include a small portion of igneous and low-grade metamorphic clasts derived from the Omineca crystalline belt along the west side of the Rocky Mountain Trench, 220 km west of the British Columbia–Alberta border (Mathews 1978; Bobrowsky 1989; Edwards and Scafe 1996; Erdmer and Cui 2009). Late Wisconsinan continental till blankets the area and directly overlies the Clear Hills/Halverson Ridge and Grimshaw gravel units. However, the Old Fort and Shaftesbury units, which occupy incised valleys (the Old Fort and Shaftesbury buried valleys, respectively), are each conformably overlain by bedded to laminated sand, silt, and clay units (Fig. 7). Dropstones and diamictic pods occur in upper parts of these units below the base of till (Mathews 1978; Balzer 2000; Leslie and Fenton 2001; Hartman and Clague 2008; Morgan et al. 2012).

Deep bedrock depressions adjacent to the mountain front near Lynx Creek are infilled with Cordilleran-sourced diamictons, lacustrine rhythmites, and debris flow deposits (Campbell 1959; Mathews 1978; Hartman and Clague 2008; Hartman et al. 2018; Fig. 5). The infill is truncated by the Shaftesbury gravel, which is in turn overlain by proglacial lacustrine deposits, montane till, and continental till (Hartman et al. 2018). Montane outwash occurs in tributaries of the Shaftesbury Valley near Lynx Creek (Mathews 1978; Hartman et al. 2018).

Four basal gravel deposits, each occupying a distinct stratigraphic and physiographic position, collectively record at least 175 km of lateral migration and 690 m of incision by the ancestral Peace River. The oldest gravel, capping the Clear Hills and Halverson Ridge, was deposited as an extensive fluvial pediment and may be mid- to late-Tertiary in age (Mathews 1978). The Grimshaw gravel was deposited on a broad strath that is roughly equivalent in height to the modern plains of the Peace River lowland. The Grimshaw strath was subsequently incised to successively lower levels by the evolving ancestral Peace River, resulting in the sequential deposition of the Old Fort and Shaftesbury gravel units on a series of step-like bedrock straths (Fig. 7) prior to and during the Middle Wisconsinan, respectively (Slomka et al. 2018).

Sedimentological evidence indicates that the only time continental ice unequivocally extended across the Peace River region was during the Late Wisconsinan. However, large, steeply dipping clinoforms, interpreted as deltaic foresets (Slomka and Hartman 2019) in the Grimshaw gravel and the upward fining of the Old Fort gravel, indicate that successive continental ice sheets advanced at least into the lower reaches of the ancestral Peace River valley prior to the deposition of Shaftesbury gravel in the Middle Wisconsinan (MIS 3). Upward fining of the Old Fort gravel likely relates to regional influence of continental glaciation during the Early Wisconsinan (MIS 4), and the sedimentary architecture of the Grimshaw gravel indicates that continental ice may have approached the region and blocked drainage even earlier, possibly prior to or during the Illinoian (MIS 6, 8). The discovery of magnetically reversed continental till on the Buffalo Head Hills (Andriashek and Barendregt 2017; Fig. 1) indicates that continental glaciation reached north-central Alberta prior to MIS 20. The distribution of pink–red granite gneiss erratics within the Old Fort and Shaftesbury gravel units as far west as the AB–BC border and Lynx Creek, respectively (Mathews 1978; Hartman and Clague 2008; Fig. 5), is consistent with the dispersal patterns of continental ice advancing southwest across the Shield, although, to date, an Omineca source for some of these erratics cannot be ruled out.

The occurrence of Cordilleran-sourced diamictons and outwash deposits that both underlie and overlie gravel infilling the Shaftesbury valley indicates that montane ice reached the western edge of the CIP prior to the Middle Wisconsinan, as well as during Late Wisconsinan glaciation (Campbell 1959; Mathews 1978; Hartman and Clague 2008; Hartman et al. 2018).

The SAOS area lies along the eastern border of Alberta (Fig. 1). As a result of drilling related to widespread resource extraction, this area has the greatest density of high-quality shallow subsurface information in the province, which includes boreholes with geophysical logs or geotechnical descriptions, core, and logged water wells. These data have enabled the identification of deeply incised buried valleys (Fig. 8) that have no physiographic expression on the modern land surface. Infill comprises basal gravel overlain by thick sequences of tills interstratified with (glacio-) fluvial and lacustrine sediments (Figs. 9 and 10). Based on stratigraphic work by Andriashek and Fenton (1989), all basal gravel deposits were initially assigned to the Empress Formation (Andriashek 2003; Andriashek and Barendregt 2017; Utting 2021).

Three main buried valleys occur in the SAOS area; the Wiau, Leismer, and Christina channels (Fig. 8). The widest is the east–west-oriented Wiau channel, which ranges from 13 to 25 km in width. The floor of this steep-walled, generally straight valley descends eastwards from 440 m a.s.l. in the western limit of the study area to 390 m a.s.l. at the AB–SK border and is infilled by sediment ≤300 m thick. The Wiau channel is paralleled to the north by the Christina channel, the floor of which occupies a similar elevation range (430–410 m a.s.l.) but is narrower (5–8 km wide) and less deeply buried (200 m of sediment). The Leismer channel, which is oriented generally north to south, (Fig. 8) ranges from 5 to 15 km in width, with a floor that descends southwards from 500 to 440 m a.s.l., where it joins the Wiau channel. Sediment thickness along the Leismer channel ranges from 175 to 200 m.

Basal gravel in the Wiau and Leismer channels (unit 1 of the Empress Formation) is 20–50 m thick, and conformably overlain by 5–20 m of silt and clay (unit 2 of the Empress Formation) and capped by 2–7 m of sand and gravel (unit 3; Utting 2021). There are hundreds of geophysical logs from oil and gas wells to delineate the geometry of these units, revealing that they form a continuous deposit over a large area. However, assessment of gravel petrology is based on just two sample locations, including drilling grab samples from WR99-1 (Andriashek 2003) in the western Wiau channel and core recovered from the Statoil OBS borehole in the Leismer channel (Fig. 8). Grab samples from WR99-1 reveal trace/minor pink–red granitic fragments in unit 1, thought to be a result of caving into the drill hole from material stratigraphically above rather than from the sediment itself (Andriashek 2003). However, the core from the Leismer channel contains a cobble-size clast of pink gneiss within unit 1, 10 m above its base (see supplementary information). This sample suggests that Shield erratics are not completely absent in unit 1 as previously thought by Andriashek (2003), or that the lower portion is Shield erratic-free while the upper portion of unit 1 contains these erratics.

The interfluve between the Wiau and Christina channels (Fig. 8) lies at ∼470 m a.s.l. and is buried beneath up to 175 m of sediment. Core WEPA000-3 (Fig. 8) recovered 25 m of glacial silt, sand, and gravel containing abundant pink–red granite gneiss clasts attributed to the Empress Formation (undifferentiated), overlain by till and fluvial deposits (Andriashek 2003). Basal gravel on the plains adjacent to buried valleys (undifferentiated Empress Formation; Andriashek 2003) has not been recovered. These units are overlain by thinner till and fluvial deposits (Figs. 9 and 10).

Basal gravel of Empress Formation unit 1 along the Wiau and Christina channels are overlain by unit 2, a fine-grained silt/clay deposit that contains sand-size fragments of pink granite presumed to be derived from the Shield. The unit is overlain by unit 3, a gravel deposit containing abundant pink to red granite gneiss fragments. These deposits, as well as the basal gravel on the interfluve and adjacent plains, are covered by the Bronson Lake Formation till (Figs. 9 and 10). Andriashek and Barendregt (2017) determined that this till is magnetically reversed, indicating that earliest glaciation in this region occurred prior to 0.78 Ma (MIS 20), with the underlying Empress Formation sediments being deposited some time earlier. Although the Bronson Lake Formation was initially considered to be restricted to the Wiau and Christina channels (Andriashek 2003), recent examinations of the lower till in WEPA 99–1 (Fig. 8; Andriashek and Barendregt 2017; Utting 2021) show that it is extensive, covering all basal gravel units in the region, including those on the adjacent plains.

The bedrock topography and stratigraphy of the SAOS area reveal an initial period of fluvial planation followed by incision and deposition of basal gravel along the floors of the Wiau, Christina, and Leismer channels. All buried valley basal gravel units (Empress Formation unit 1) are interpreted as fluvial deposits that were emplaced in the Early Pleistocene (pre-MIS 20; >0.78 Ma). Units 2 and 3 are interpreted as increasingly ice-proximal glaciolacustrine to glaciofluvial sediments deposited, as advancing continental ice initially blocked regional drainage, impounding a proglacial lake. The gneiss clast within unit 1 of the Leismer channel core may indicate a transition to glacial or ice-proximal conditions. These sediments were then covered by Bronson Lake Formation till as a pre-Illinoian continental ice sheet extended across the area (Andriashek 2003). The sand and gravel deposits on the interfluve had previously been considered preglacial (Stein et al. 1993). However, the recovery of Shield erratics from this gravel prompted reinterpretation as a glaciofluvial deposit (Andriashek 2003).

Basal gravel on the plains adjacent to buried valleys has not been recovered and thus its petrological composition has not been determined. However, based on a pattern of fluvial incision and planation, the basal gravel on the plains is older than the gravel (unit 1) along the base of buried valleys, while the interfluve gravel would be intermediate in age. However, the glaciofluvial genesis of the interfluve gravel suggests that it post-dates the preglacial basal gravel within the Wiau and Leismer channels. We note that assessment of the petrologic composition of the Wiau and Leismer basal gravel units is based on very few samples; therefore, unit 1 may not be Shield clast-free and may post-date the interfluve gravel. Alternatively, the higher physiographic position of the interfluve gravel may be due to valley filling and lateral planation, and/or a subsequent glacial configuration that has not yet been described.

In the Peace River area, three successive basal gravel units record fluvial incision, planation, and deposition between successively more extensive continental glaciations advancing from the east and two montane glaciations advancing from the west. The elevation of the modern Peace River, 50 m below the floor of the Middle Wisconsinan paleovalley, indicates that incision continued in the Holocene. Within the SAOS region, preglacial basal gravel underlies a succession of five glacial and interglacial units, of which the lowest till (Bronson Lake Formation) is magnetically reversed indicating that it, and the underlying basal gravel, were deposited in the Early Pleistocene (pre-MIS 20). The uppermost interglacial unit (the Sand River Formation) was deposited by rivers active during the Middle Wisconsinan (MIS 3), and the overlying Grande Centre Formation till was deposited by the Late Wisconsinan (MIS 2) continental ice sheet.

Infill of the SMW buried valley system includes multiple fluvial gravel units deposited prior to, during, and between glaciations. However, only the lowest units immediately underlying the first glaciogenic sediments containing Shield erratics satisfy the preglacial requirements of the current Empress Group definition (Whitaker and Christiansen 1972; Evans and Campbell 1995; Jackson et al. 2011). These requirements are problematic when considering, for example, the Shaftesbury and Old Fort gravel units. The Old Fort gravel, which is Shield clast-free west of the B.C.–Alberta border as well as immediately preglacial, cannot be assigned to the current Empress Group because it occupies a higher strath adjacent to that which underlies the Shaftesbury gravel. However, the Shaftesbury gravel can only be considered lithologically preglacial west of sites where Shield erratics are absent and east of where montane outwash is reported, which are both near Lynx Creek (Figs. 1 and 5; Mathews 1978; Hartman and Clague 2008; Hartman et al. 2018). Therefore, only a small portion of the Shaftesbury gravel satisfies the current conditions for Empress Group assignment.

While absolute dates are unavailable from most of the units infilling the SMW buried valley system, sedimentological evidence clearly demonstrates that its eastern reaches evolved significantly earlier than in the west. This is emphasized by the stratigraphic position of Middle Wisconsinan deposits at either end of the SMW system, which in the east occur near the top of the valley fill succession (Sand River Formation; Figs. 2 and 10) and in the west, at the base (Shaftesbury gravel; Figs. 2 and 7).

Modelling results show the distribution of basal gravel across Alberta, including isolated remnants of high planation surfaces and more contiguous deposits along successive step-form bedrock straths and regionally integrated paleovalley systems (Fig. 3). These record multiple cycles of fluvial incision, lateral planation, and deposition by ancestral rivers flowing eastwards across the CIP following uplift of the Rocky Mountains during the Laramide Orogeny between 80 and 35 Ma (Roed 1968; Vonhof 1969; du Toit 1988; Osborn and du Toit 1991; Leckie 2006).

The recognition that discrete basal gravel deposits are associated with mappable straths across Alberta provides a physiographic criterion to classify previously undifferentiated units across the CIP (e.g., units associated with strath terraces and floors of buried valleys). Furthermore, buried valley systems, such as the SMW that extend across the CIP, evolved sequentially from east to west prior to and between glaciations and thus basal gravel along apparently contiguous valley floors is comprised of multiple chronostratigraphic units. It is also important to note that units occupying the same relative physiographic position (e.g., valley floor) but within different buried valley systems may not be implicitly considered stratigraphically equivalent because the incision history of each fluvial system may be partly dependent on watershed-specific characteristics (Bufe et al. 2016; Foster et al. 2017).

Future studies to refine regional Quaternary stratigraphy elsewhere in Alberta would benefit from a similar approach of modelling bedrock strath geometry and the distribution of associated basal gravel deposits (e.g., Fig. 4). Once identified, basal gravel allounits could be comprehensively mapped, described, and utilized to better understand regional geologic histories.

Except for Early Pleistocene units in the SAOS region, most basal gravel units in Alberta have been deposited between repeated montane and continental glaciations. This is exemplified by the Old Fort and Shaftesbury gravel units, where glacially transported erratics are absent along the western parts of the deposits, which would have occupied an ice-free corridor between the converging montane and continental ice sheets prior to the Middle Wisconsinan, and again during the Late Wisconsinan when the corridor finally closed (Hartman et al. 2018). Both units fine upwards into bedded sands and silt–clay rhythmites deposited in proglacial lakes that formed as continental ice extended westward, and are truncated by till deposited as the region was overridden by Late Wisconsinan continental ice. The complex glacial history of the CIP, characterized by multiple spatially and temporally transgressive advances and retreats of continental and montane ice chronostratigraphically intercalated between successive fluvial incision and planation events renders the preglacial requirement of the current Empress Group definition unworkable, since mappable unconformity-bounded basal gravel units may be preglacial only in reaches upstream of previous glacial limits. However, recognition of basal gravel as immediately preglacial, recorded by conformably overlying glaciogenic fine-grained strata, conveys information regarding the stratigraphic positions of basal gravel and overlying till. Therefore, we suggest the requirement that Empress Group is immediately preglacial relative to the first-ever advance of continental glaciation across the CIP be replaced with the requirement that Empress Group units are conformably overlain by the earliest glaciogenic sediments of continental ice sheet origin, regardless of the age of the associated glacial advance across a specific area.

Although basal gravel deposits along the floors and strath terraces of buried valleys in Alberta are not stratigraphically equivalent, we recognize that such units are sequential and related and may be grouped (see Article 28 of the North American Stratigraphic Code (NACSN 2021)). Whereas all products of lateral planation by ancestral fluvial systems might be included in a single group, irrespective of their physiographic position across the CIP, the utility of grouping basal gravel deposits based on lithology (fining upward), stratigraphic position between bedrock and till, and physiographic setting within integrated buried valley networks is that it provides a criterion to map and classify fluvial units that is meaningful to geologists and non-geologists (McMillan 2005; Salvador 2013) and builds on well-established Empress Group terminology (Whitaker and Christiansen 1972; Cummings et al. 2012; Fig. 2). However, without formation-level stratigraphic definitions, such grouping conveys little stratigraphic knowledge, particularly regarding the relationship between basal gravel units and age and extent of overlying tills that span successive glaciations, beyond the fact that coarse-grained fluvial sediments are widely distributed and occupy numerous stratigraphic positions within buried valleys.

Formation status has been assigned to only a few buried valley basal gravel units within the CIP, including those in the SAOS area, as well as the Saskatchewan Gravel Formation in southwestern Alberta and southern Saskatchewan (McConnell 1885; Jackson et al. 2008). Elsewhere, basal gravel along the floors of paleovalley networks have typically been assigned to the Empress Group without establishing constituent formations (Whitaker and Christiansen 1972), or are only recognized informally, despite having been thoroughly described (e.g., basal gravel units in the Peace River region; Slomka et al. 2018; Slomka and Hartman 2019). This approach is procedurally backward (Salvador 2013) and obscures the stratigraphic resolution that formal definitions of basal gravel formations may provide.

By combining lithostratigraphy with allostratigraphic elements in the Peace River region, we propose that the Old Fort and Shaftsbury gravel units, which have previously been used as informal stratigraphic names (Slomka et al. 2018; Slomka and Hartman 2019), be recognized as the Old Fort and Unchaga Formations, respectively, of the Empress Group. The criteria for formal recognition of the Old Fort and Unchaga formations following the North American Stratigraphic Code (NACSN 2021) are included in the accompanying supplementary information. The Shaftsbury gravel is named the Unchaga Formation, meaning “large river,” the Dane-Zaa name for Peace River (Leonard 2018), because the Shaftsbury Formation name is currently reserved for the Lower Cretaceous shale underlying much of the Shaftesbury strath (Glass 1990; Prior et al. 2013). Both the Old Fort and Unchaga formations are included in the Empress Group because they occupy discrete bedrock straths within a single buried valley, extending approximately 400 km through the Peace River region.

Proglacial lacustrine deposits that conformably overlie both the Old Fort and Unchaga formations may also be included as discrete members within these formations (e.g., the Glacial Lake Mathews unit; Hartman 2005). Basal gravel on the Clear Hills/Halverson Ridge and the Grimshaw gravel would not be included in the Empress Group because they are unconformably overlain by glaciogenic sediments containing Shield erratics and are not situated in buried valleys; however, both may be defined as discrete formations (cf. Slomka et al. 2018; Slomka and Hartman 2019). Glaciofluvial or post-glacial units that occupy the same position overlying bedrock and beneath till such as tunnel valley deposits (Atkinson et al. 2013) are separated from basal gravel units by an unconformity and would be excluded from the group.

In the SAOS region, we advocate that the Empress Formation (Andriashek and Fenton 1989) be renamed the Ipiatik Formation (see supplementary information) because use of the same name at different stratigraphic ranks is forbidden (see Article 7 of the North American Stratigraphic Code (NACSN 2021)). The Empress Group (Whitaker and Christiansen 1972) must be retained because it has precedence. Constituent units 1, 2, and 3 of the Empress Formation (Andriashek 2001, 2003; Parks and Andriashek 2002; Utting and Andriashek 2020; Utting 2021; Fig. 1) are defined as the House, Wappau, and Calder members, respectively (see supplementary information). We also advocate for recognition of the interfluve gravel as the Winefred Formation. A definition of the formation is provided in the supplementary information.

Basal gravel deposits are the terrestrial sedimentary record of post-orogenic fluvial exhumation of the CIP that began in the Early Miocene and continues to the present. These deposits occupy distinct stratigraphic and physiographic positions that document successive cycles of fluvial incision, planation, and deposition. The most complex sedimentary successions occur within the buried valleys of Saskatchewan and northeastern Alberta, which are floored by preglacial gravel deposited during the Early Pleistocene (Barendregt et al. 2012; Andriashek and Barendregt 2017). However, basal gravel in central and western Alberta, such as those in the Shaftesbury buried valley, were deposited more recently. Although buried valleys appear contiguous, they developed sequentially from east to west, with basal gravel units becoming progressively younger to the west with each successive glaciation.

By combining the results of new modelling showing the distribution of discrete basal gravel deposits on straths (bedrock unconformities) with traditional lithostratigraphy, we have proposed new stratigraphic frameworks for the fluvial successions of the Peace River and SAOS areas whereby the Old Fort, Unchaga, Ipiatik, and Winefred formations are included in the Empress Group. In doing so, we emphasize the continued utility of the currently defined Empress Group, albeit with the minor modification that its requirement to be the youngest preglacial gravel in an area be removed due to the absence of a single isochronous glacial unconformity. Instead, the Empress Group now includes all sequentially related gravel deposits and conformably overlying fine-grained glaciogenic deposits that immediately predate any glaciation of an area and occupy bedrock-floored buried valleys, regardless of whether they occur along the valley floors or adjacent strath terraces. We suggest that the approach and techniques presented in the paper be applied elsewhere in the province, where to date, constituent formations of the Empress Group have not been defined. Such work will provide further improvements to understanding the geologic history of this important Quaternary unit in the CIP.

We thank Associate Editor Roger Paulen and two anonymous reviewers for constructive feedback that improved the manuscript. We also thank our colleagues Dr. Matt Grobe and Dan Palombi who reviewed early drafts of the manuscript. Importantly, we acknowledge the significant contributions of Laurence Andriashek to the Quaternary stratigraphy of northeast Alberta, and for the many insights he has shared on the age and origin of preglacial gravel throughout the region. Finally, we thank Elder Fred Didzena of the Dene Tha’ First Nation, and our colleague Brain Lieverse for considering and supporting our use of the name “Unchaga.” This work has been supported by Alberta Innovates through their Water Innovation Program.

All data that support the findings of this submission are previously published and cited accordingly.

Conceptualization: GMDH, SMP, DJU, NA, JEL

Data curation: GMDH, SMP, DJU, NA

Formal analysis: GMDH, SMP, DJU, NA

Investigation: GMDH, SMP, DJU, NA

Methodology: GMDH, SMP, DJU, NA

Project administration: JEL

Resources: JEL

Supervision: JEL

Validation: GMDH, SMP, DJU, NA

Visualization: GMDH, SMP, DJU, NA

Writing – original draft: GMDH, SMP, DJU, NA

Writing – review & editing: GMDH, SMP, DJU, NA, JEL

There is no funding to report for this submission.

Supplementary data are available with the article at https://doi.org/10.1139/cjes-2022-0143.