High‐frequency sequences within a retrogradational deltaic succession: Upper Cenomanian Dunvegan Formation, Western Canada Foreland Basin

After prograding for several hundred kilometres during Middle Cenomanian time, the Dunvegan delta complex in north‐west Alberta and adjacent British Columbia experienced stepwise transgression, commencing at about the Middle to Late Cenomanian boundary. Progressive drowning of the delta complex is recorded by Dunvegan allomembers B and A, each comprised of three simple depositional sequences, bounded by composite subaerial unconformity/flooding surfaces. Each sequence represents an array of deltaic depositional environments. Delta‐front sandstones preserve little evidence, such as hummocky cross‐stratification, for powerful storm wave action, although wave and combined‐flow ripples are common. Delta‐front sandstone bodies tend to be smaller and lobate in the lower part of the studied interval, and larger and more linear near the top. This suggests increasingly effective wave‐driven redistribution of sand as more open‐marine conditions were gradually established. The top surfaces of allomembers B and A are locally incised by sandstone‐filled palaeovalleys up to 19 m deep; river incision may have been a response to relative sea‐level fall and/or a change in the ratio of discharge to sediment load. Overall, the shoreline described a broad arc, open to the south east, with the sense of shoreline migration north‐west to south‐east. For each sequence, the shoreline migrated an average of 80 km between transgressive and regressive limits. The transgressive limit shows a progressive landward offset of about 15 km per sequence, culminating in complete drowning of the delta system above sequence A3. Isopach maps show that syn‐depositional tectonic subsidence rotated the basin down to the south‐west; palaeogeographic maps show, however, that the sea floor sloped to the south‐east, implying that sediment redistribution effectively filled all tectonically generated accommodation and maintained a south‐east inclined depositional surface. Transgressions and regressions across this surface were therefore driven primarily by eustasy rather than pulses of tectonic subsidence. Simple calculations based on inferred alluvial gradients of 10–20 cm/km suggest that eustatic excursions of ca 8–16 m would have been sufficient to generate sequence thicknesses on the order of 10 m. Limited geochronologic and biostratigraphic control suggests that the six simple sequences that form Dunvegan allomembers B and A each represent an average of about 41 kyr, suggesting that the orbital obliquity cycle was the primary control on high‐frequency sea‐level cycles.


| High-frequency sea-level change
This paper documents a succession of high-frequency (<10 5 year timescale) depositional sequences, organized in an overall backstepping stacking pattern. The rocks record the drowning of a large, Middle to Upper Cenomanian deltaic complex in the Western Canada Foreland Basin, and show that this was not a simple, progressive transgression but involved multiple shoreline oscillations. The mechanisms responsible for generating high-frequency (10 4 -10 6 year timescale), transgressive-regressive sequences in Cretaceous shallow-marine deposits are a subject of intense debate. There is persuasive evidence that Cretaceous third and fourth (and possibly fifth and sixth)-order sequences are inter-regionally correlatable, and therefore of eustatic origin (Haq, 2014;Sames et al., 2016). The critical question is 'what drove eustatic change'? Some studies (Plint, 1991;Miller et al., 2005a;2005b;Plint and Kreitner, 2007;Kominz et al., 2008;Galeotti et al., 2009;Koch and Brenner, 2009;Uličný et al., 2009;Boulila et al., 2011;Lin et al., 2019) have argued that >10 m of sea-level change in <1 Myr can only be explained, even under 'greenhouse' conditions, by the waxing and waning of ice sheets, governed by orbitally modulated variation in insolation. The operation of a glacio-eustatic mechanism, even during Cretaceous 'greenhouse' conditions, appeared to gain support from modelling experiments (Flögel et al., 2011). Nevertheless, isotopic studies (Moriya et al., 2007;MacLeod et al., 2013) found no evidence for significant terrestrial ice accumulation, particularly during Cenomanian-Turonian 'hot-greenhouse' conditions. An alternative mechanism, initially proposed by Hay and Leslie (1990), postulated that terrestrial aquifers had the capacity to absorb and release water in response to changes in rainfall intensity, driven by Milankovitch cycles. Recent estimates suggest that, in the Cretaceous, this 'aquifer-eustasy' mechanism could have produced up to about 80 m of sea-level change on a timescale of 10 4 -10 5 years (Kidder and Worsley, 2010;Haq, 2014;Sames et al., 2016;Hay et al., 2019), and thus was as effective as glacio-eustasy. Investigations that have integrated the physical stratigraphy of mid-Cretaceous rocks with oxygen and carbon-isotope analyses Laurin et al., 2019), have shown that climate warming coincided with sea-level fall whereas sea-level rise occurred during climate cooling: This is the response predicted by the aquifer-eustasy model, and the opposite of the glacio-eustatic model.
The Middle to Upper Cenomanian deltaic sequences, here documented from the Dunvegan alloformation (allostratigraphic definition following Bhattacharya and Walker, 1991a;1991b;Plint, 2000), in the Western Interior Seaway in Canada, record sea-level oscillations of <ca 20 m on a timescale of 10 4 years. Sea-level cycles of similar magnitude and frequency have also been recognized through the overlying Upper Cenomanian lower Kaskapau alloformation (A-X, Doe Creek and Pouce Coupe units; Plint, 2019;Plint and Kreitner, 2019). These high-frequency sequences are superimposed on a longer-term marine transgression that culminated in Oceanic Anoxic Event 2, spanning the Cenomanian-Turonian boundary (Van Helmond et al., 2016).
This study, focused on the uppermost part of the Dunvegan alloformation, documents evolving palaeogeography during a pulsed, long-term transgression, and permits a first-order approximation of both the amplitude and frequency of relative sea-level cycles during this 'hot greenhouse' time. This investigation complements that of Merletti et al. (2018) who document, from the Campanian Almond Formation in Wyoming, a complex, multi-phase backstepping of the shoreline in which a protracted (1.7 Myr) transgression was punctuated by repeated minor transgressive-regressive cycles of ca 20 kyr duration. Each short-term sea-level fall was accompanied by the lateral progradation of wave-dominated barrier spits or strandplains across the mouth of a broad embayment filled with lagoonal and coastal plain facies. continental margin during Jurassic and Cretaceous time resulted in major crustal shortening, compression, and development of a broadly north-east facing fold and thrust belt (McMechan and Thompson, 1993;Price, 1994;Evenchick et al., 2007). Static loading by the fold and thrust belt, coupled with regional dynamic subsidence driven by mantle flow, resulted in the down-warping of the western margin of North America, producing a retro-arc foreland basin a few hundred kilometres wide. Palaeozoic and older Mesozoic sedimentary rocks were eroded from the ancestral Rocky Mountains to the west, and their detritus was trapped primarily in the adjacent foredeep, building a succession of deltaic and strandplain complexes. To the east of the foredeep and beyond the forebulge lay a broad, shallow epeiric seaway up to ca 1,000 km wide, subsidence of which is attributable largely to the effect of mantle flow, coupled with minor isostatic subsidence driven by sediment and water loads (Mitrovica et al., 1989;Price, 1994).

| Stratigraphic setting, terminology and database
Because of a very large and publically accessible wireline log and core database, the Western Canada Foreland Basin provides an unparalleled opportunity to investigate, on a basin scale, the stratigraphy, sedimentology and depositional history of Cretaceous shallow-marine clastic successions. Well control allows three-dimensional stratigraphic architecture and broad lithological distributions to be mapped in subsurface. Depositional sequences can be correlated from boreholes to outcrop in the Rocky Mountain Foothills, permitting facies and palaeocurrent detail to be added to the regional palaeogeographic picture determined from subsurface data. The Cenomanian Dunvegan Formation represents a huge delta complex that prograded from north-west to southeast, extending for many hundreds of kilometres along the foredeep of the foreland basin (Stott, 1982). The Dunvegan Formation, defined as a lithostratigraphic unit, gradually thins out south-eastward as sandstone-rich alluvial and delta-front facies grade downdip into coeval prodelta mudstones, the latter assigned, in different parts of the basin, to the lithostratigraphic Cruiser, Shaftesbury or Blackstone formations (Stott, 1982). The downdip passage of sandstone into mudstone makes it difficult to establish temporal relationships between the various facies regions, and hence determine the depositional and palaeogeographic history of the formation.
The Dunvegan allomembers were originally defined, on pragmatic grounds, as rock packages bounded by regionally mappable flooding surfaces, most of which merged updip into subaerial unconformities incised by valleys. Allomembers are, however, composite rock bodies, composed of several transgressive-regressive successions (the 'shingles' of Walker, 1991a, andPlint, 1996), some of which embody evidence of both relative rise and fall of sea level: The latter is commonly manifest as sharp-based shoreface sandstone bodies of the Falling Stage Systems Tract (Plint and Nummedal, 2000). The boundaries of 'shingles' are not, however, incised by valleys. Thus the upward-shoaling successions that form 'shingles' could be interpreted either as simple sequences if they embody evidence for deposition through a full cycle of sea level, or parasequences if they appear to record only pulses of relative sea-level rise (Van Wagoner et al., 1988). Because most of the six upward-shoaling successions that form Dunvegan allomembers B and A are capped, at least locally, by sharp-based delta-front sandstones, these successions will be referred to as simple, Exxon-type sequences, deposited during both relative sea-level rise and fall.
The detailed stratigraphy of allomembers B and A, that form the uppermost part of the Dunvegan alloformation, was never fully resolved by either Bhattacharya (1989) or Plint (2000). This problem was addressed by Hay (2006), who based his analysis on a data set of 970 gamma ray and resistivity well logs and 15 cores distributed over about 30,000 km 2 of the Alberta and British Columbia Plains, supplemented by 23 outcrop sections in the Rocky Mountain Foothills and Peace River Valley ( Figure 1). Bounding surfaces, generally comprising composite subaerial unconformities and marine flooding surfaces, were correlated and looped through a grid of 37 working cross-sections. The resultant stratigraphic framework (Hay, 2006) forms the core of the present study. Subsequent work by Plint used an additional 655 well logs to extend the mapping of allomembers B and A eastward for a further 250 km in order to determine their distal limits, as summarized by Hay and Plint (2009

| Dunvegan allomembers B and A
Allomember C of the Dunvegan Formation records the maximum regression of the delta complex (Bhattacharya | 527 HAY And PLInT and Walker, 1991a;1991b;Plint, 2000). The overlying allomembers B and A record the progressive backstep of the shoreline towards the north-west in response to long-term relative sea-level rise (Bhattacharya and Walker, 1991a;1991b;Hay and Plint, 2009). In the original allostratigraphic scheme of Bhattacharya and Walker (1991a), allomember B was defined as a single upward-shoaling succession (which was termed a 'shingle' by those authors), whereas allomember A comprised two shingles. Further work by Hay (2006) showed that six mappable upwardshoaling successions could be recognized between the top of Dunvegan allomember C and the top of allomember A. A major subaerial unconformity capping the third sequence was recognized by Hay (2006). In consequence, allomember B was re-defined to include the three sequences (B1-B3) below the subaerial unconformity whereas allomember A comprised the three sequences (A1-A3) above ; Figure 2). The erosion surfaces that bound sequences B1-B3 and A1-A3 generally lack a coarse-grained lag and provide little evidence for deep scour into the underlying sediment. An exception is provided by the top surface of allomember B which is incised by sandstone-filled palaeovalleys. Sparse core and outcrop control shows that, between palaeovalleys, allomember B is capped, at least locally, by a thick palaeosol that is interpreted to represent an interfluve surface. Rare palaeovalley fills are also known at the top of allomember A ).

| FACIES
The sedimentary facies and ichnology of the more seaward portions of the Dunvegan delta complex were described by Bhattacharya and Walker (1991b), Bhattacharya (1993), Plint (1996), Gingras et al. (1998), Coates and MacEachern F I G U R E 1 Map of study area showing location of well logs, cores and outcrop sections. Core cross-sections C1, C2 and C3 are shown in red and outcrop cross-sections O1, O2 and O3 are shown in green. Inset map shows study area in relation to Alberta and British Columbia. The pale green line shows the approximate position of the regional dip section shown in Figure 2A (2007), and Bhattacharya and MacEachern (2009), whereas more updip, coastal plain facies and environments were documented by McCarthy and Plint (1998;2003);, Plint et al. (2001);McCarthy (2002); Plint and Wadsworth (2003) and Lumsdon-West and . The sedimentary facies of Dunvegan allomembers B and A have, to a large extent, been documented in these previous studies, and hence detailed description need not be reiterated. The principal facies and interpreted depositional environments of allomembers B and A are summarized in Table 1, and illustrated in Figures 3 through 7.

| Facies successions in core and outcrop
The regional sequence stratigraphic framework of the delta complex was established through detailed wireline log correlations, summarized in Hay and Plint (2009). Interpretation of the various depositional environments represented by Dunvegan allomembers B and A must, however, be based on observation of core and outcrop. To that end, six regional cross-sections, located in Figure 1, summarize the facies observed in subsurface (Figures 8 and  9), and in outcrop (Figures 10 and 11). It is important to appreciate that correlation between outcrop sections was guided by the regional subsurface allostratigraphic framework, in which transgressive marine mudstone blankets formed the primary stratigraphic markers. In contrast, sandstone-rich units typically display rapid lateral change in facies and thickness which makes them impossible to correlate over long distances. The outcrop and core-based stratigraphy in Figures 8 through 11 is supplemented by traces of resistivity logs that help to illustrate the character and boundaries of the six depositional sequences B1-B3 and A1-A3.
The facies described in Table 1 and illustrated in Figures 3 through 11, can be grouped into five distinct F I G U R E 2 (A) Summary dip section through the Dunvegan alloformation showing the stacking pattern of allomembers J to A, and the broad distribution of delta plain alluvium, delta-front sandstone and prodelta to offshore mudstone. Allomembers J through C are organized in an overall south-east directed, progradational stacking pattern, whereas allomembers A and B backstep to the north west. Dunvegan allomembers downlap onto a regional, highly phosphatic and radioactive condensed section (FSU marker) that forms the uppermost part of the Fish Scales alloformation (Roca et al. 2008). The base of the Kaskapau alloformation marks a major shoreline backstep, after which deltaic facies are replaced by more open-marine facies that form the basal Kaskapau A-X unit (Plint, 2019). (B) Detailed stratigraphic organization of allomembers A and B, each of which consists of three simple depositional sequences, bounded by regional flooding surfaces that pass updip into subaerial unconformities, some of which are incised by valley systems. The six sequences that form allomembers A and B are organized in an overall backstepping stacking pattern T A B L E 1 Summary of the principal characteristics, and palaeoenvironmental interpretation of the facies that form Dunvegan allomembers B and A BI = 4-6 Prodelta, mainly low energy but subject to flood-or storm-driven sand influx. Near-normal marine salinity, relatively low rate of deposition with pervasive bioturbation 8 Very fine to fine-grained sandstone with 10%-50% dispersed mud; commonly structureless due to intense bioturbation; some parallel and cross-lamination preserved locally. Rare siderite nodules with clay ooids ( Figure 6A vertical facies successions, designated A-E ( Figure 12). Facies successions represent distinct suites of sub-environments within the delta complex (cf. Elliott, 1974;Bhattacharya, 1993).

| Succession A
Facies succession A (Figure 12), typically begins with laminated to moderately bioturbated dark mudstone with Lingula and dispersed fish scales (facies 1). Mudstone grades up into inter-bedded wave-rippled sandstone and mudstone (facies 5) or, rarely, current-rippled heterolithic sandstone (facies 6). The Bioturbation Index (BI) in these heterolithic facies is low (BI 0-2). The rippled heterolithic facies 5 may be capped by a sharp transgressive surface marking the base of the next succession, or a palaeosol of facies 3, or rarely, by an erosive-based distributary channel sandstone of facies 11. The entire succession may reach a maximum thickness of 10 m, although 5-6 m is more usual; succession A is capped by a marine transgressive surface.

| Interpretation
Succession A is interpreted to represent prodelta and delta-front deposits of a wave-influenced deltaic shoreline. The scarcity of unidirectional current ripples suggests that sand was not supplied directly by hyperpycnal flows issuing from river mouths, although it is possible that hyperpycnal flows were generated by storm wave resuspension (cf. Bhattacharya and MacEachern, 2009;Plint, 2014). The low intensity of bioturbation may be evidence for deposition on the down-drift side of a deltaic promontory where benthic fauna were inhibited by a high turbidity and low-salinity hypopycnal plume (Hampson and Howell, 2005;MacEachern et al., 2005). The scarcity of hummocky cross-stratified (HCS) and swaley cross-stratified (SCS) sandstone suggests that these deltas were rarely affected by large storm waves. This contrasts with deltaic successions described by Bhattacharya and Walker (1991b) and Plint (1996) from the lower part of the Dunvegan Formation where HCS and SCS were observed in successions 15-20 m thick. It seems likely that the relatively thin (<10 m) facies successions present in allomembers B and A represent deposition on a wide, shallow platform across which storm waves were damped before reaching shore. Cross-bedded sandstones at the top of successions probably represent the fill of distributary channels and the veneer of plant-rich and dinoturbated mudstone represents subaerial channel-margin or delta plain deposits. Intensely bioturbated silty fine-grained sandstone, abundant bivalves, wood debris, rare wave ripples ( Figure 6D)

| Succession B
Facies succession B (Figure 12), up to 12 m thick, begins with 1-2 m of laminated to moderately bioturbated mudstone (facies 1), that grades up through 1-2 m of silty sandy mudstone (facies 7), to muddy sandstone (facies 8), all intensely bioturbated by a Cruziana ichnofauna. In some instances, the bioturbated facies are erosively capped by up to 7 m of stratified, fine-grained sandstone of facies 10. Regional mapping shows that the stratified sandstone forms lobate to elongate bodies, typically tens of kilometres in length and a few kilometres in width. The clean sandstone of facies 10 is dominated by a mixture of horizontal and low-angle (? HCS) lamination, and trough cross-stratification in sets up to ca 30 cm thick. Thin intervals of preserved wave ripples may also be present.
Ophiomorpha may be distributed throughout the sandstone, and tends to be common below the eroded upper flooding surface where sandstone is overlain by offshore mudstone of facies 1.

| Interpretation
The intense bioturbation throughout the lower part of succession B suggests a distal delta-front setting subject to little salinity, turbidity, or oxygen stress, and supporting a Cruziana ichnofauna (MacEachern et al., 2005). Alternatively, bioturbated facies 7 and 8 might represent a more proximal prodelta setting located up-drift from a distributary, where seawater was undiluted and sediment supply rate was relatively low (Bhattacharya and Giosan, 2003;MacEachern et al., 2005). The sandier-upward succession coupled with the lobate to elongate shape of the sandstone bodies (discussed below) suggests deposition in a prograding deltaic shoreline. The abundance of planar and low-angle lamination in facies 10 suggests significant wave influence on the delta front; trough cross-stratification may represent dunes migrating in the surf zone, or possibly may record river efflux in distributary channels.

| Interpretation
The sandier-upward succession, with predominantly currentrippled sandstone interbeds suggests predominantly unidirectional, perhaps hyperpycnal flows, possibly related to seasonal floods. Depending upon local thickness and lateral extent, succession C might represent both larger distributary mouth bars building into the open sea, and also smaller crevasse deltas filling inter-distributary bays or delta-top lakes.
The abundant bands of siderite in the more offshore mudstone facies may record salinity fluctuating between fresh and marine (Bhattacharya and MacEachern, 2009). Evidence of reduced salinity is also indicated by the low-diversity molluscan fauna. The palaeosols of facies 3 are comparable to the poorly drained palaeosols described by Lumsdon-West and  and provide evidence of waterlogged conditions on the delta plain.

| Succession D
Facies succession D ( Figure 12) has only been observed at outcrop in the more updip parts of the delta complex exposed along the Peace River. The succession, typically 3 m to a maximum of 5 m thick, comprises a lower 1-4 m of sideritic laminated mudstone rich in plant debris (facies 2). This may be overlain by less than 1 m of current-rippled or structureless fine sandstone with roots and dinoturbation (facies 6). Regardless of the presence or absence of sandstone, the upper part of the succession is formed by up to 1 m of pale grey, rubbly rooted mudstone (facies 3).

| Interpretation
The laminated sideritic mudstone that forms the bulk of this succession is interpreted to have been deposited in an interdistributary bay or delta-plain lake to which large volumes of comminuted plant debris were introduced (McCarthy and . The small volume of sandstone present in the succession may have been introduced by crevasse splays or particularly large floods. The upward transition to poorlydrained palaeosol records the gradual filling of the bay and intermittent emergence. The restriction of this essentially freshwater facies to more northerly updip areas is consistent with the overall palaeogeography of allomembers A and B, which assume a progressively more terrestrial aspect towards the north and west (Bhattacharya and Walker, 1991b;Plint, 2000;Hay, 2006;Hay and Plint, 2009).

| Succession E
Facies succession E (Figure 12) is always underlain by a rooted palaeosol (facies 3) overlain, at an initial flooding surface, by a coal or carbonaceous mudstone (facies 4). This is sharply overlain by laminated to structureless mudstone (facies 2; <50 cm thick) that grades up into bioturbated silty sandstone containing abundant brackish-water bivalves (Brachydontes, Corbula, Ostrea), lumps of wood and fine plant debris (facies 9; ca 1 m thick). The intensity of bioturbation generally increases upward; recognizable forms are primarily Skolithos. Rare wave and current ripples are present near the top of the succession, although roots are never observed. The sandier-upward succession is sharply overlain by laminated mudstone of facies 2, or more rarely, facies 1.

| Interpretation
Succession E records a rising base level, with a poorlydrained palaeosol giving way to marsh or peat mire, followed by submergence beneath shallow brackish water of an inter-distributary bay. The abundance of bivalves in a bioturbated silty sandstone suggests a relatively low sedimentation rate, favourable to filter-feeding bivalves and burrowing infauna. The presence of marine mudstone (facies 1) above succession E indicates a transition to a more open shelf following marine transgression. Alternatively, the presence of facies 2 at the top of the succession suggests development of another inter-distributary bay, perhaps related to delta lobe switching.

| Sandstone isolith and palaeogeographic maps
The transgressive limits of marine mudstone and the regressive limits of delta-front sandstone were mapped F I G U R E 7 (A) Facies 10, fine-grained, sharp-based (arrow) sandstone dominated by low-angle cross-lamination, rare cross-bedding and ripple cross-lamination. Upper 2.8 m of sandbody is heterolithic with wave ripples, cross and planar lamination; B.I. may reach 4-5. Upper 50 cm of sandbody is wave-rippled and intensely deformed-possibly dinoturbated (DTB). This succession may represent the fill of a distributary channel that became abandoned, accumulating more muddy sediment at the top and eventually becoming emergent; 6-7-65-1W6, 1,617-1,625.2 m, each core sleeve = 75 cm. (B) Facies 11, mainly cross-bedded fine-grained sandstone resting erosively (lower arrow) on bioturbated muddy sandstone of facies 8. Rare Ophiomorpha are present throughout sandbody but become common in upper 30 cm. Upper arrow marks contact with coally mudstone (facies 4), overlain at transgressive surface (white line), by bioturbated sandy mudstone (facies 7). Sandstone interpreted to represent a distributary mouth bar with some marine influence;10-15-71-11W6, 1,276-1,277.5 m, each core sleeve = 60 cm throughout the regional grid of wireline well logs ). The thickness of interpreted 'clean' delta-front sandstone was measured from gamma ray logs at a cut-off value of ca 80 API. The distribution of 'clean' sandstone is portrayed in isolith maps for the six sequences that comprise allomembers B and A (Figures 13  through 15). In addition to sandstone thickness, Figures 13  through 15 show the transgressive limit of marine mudstone (identified where a simple marine mudstone log signature passes laterally into heterolithic 'spiky' coastal plain facies), the regressive limit of delta-front sandstone (approximated by the 1 m sandstone isolith), the type of facies succession (A-E) present in each core or outcrop section, and palaeocurrent data measured at outcrop. The data in Figures 13 through 15 formed the basis for interpretive palaeogeographic block diagrams (Figure 16), and a dip-oriented Wheeler (chronostratigraphic) diagram (Figure 17), that also includes data from Hay and Plint (2009), with supporting information on allomember C from Bhattacharya (1989).

| Sequence B1
The transgressive limit of sequence B1 (Figures 13, 16 and  17), is displaced about 210 km landward (north-west), measured perpendicular to the regressive limit of the underlying allomember C delta-front sandstone. This transgressive event marks the change from net progradation of allomembers J to C, to net backstep of allomembers B and A, that culminated in complete drowning of the Dunvegan delta system (Figure 2). Updip from the marine transgressive limit of B1, facies succession D, representing brackish bays and lakes, is dominant, whereas to seaward, facies successions A, B and C predominate, indicating a more open marine setting. Wave ripple crests tend to parallel the inferred palaeoshoreline whereas current ripples indicate broadly south easterly flow towards the coast. Patchy sandstones located landward of the marine transgressive limit may represent mouth bars in broad, shallow bays and lakes on the delta plain; deposition probably took place during base-level rise.

F I G U R E 8
Strike-oriented core cross-section C1 linking to outcrop on Wapiti River, located in Figure 1 | 537

HAY And PLInT
The regressive limit of the delta front is readily mapped in the south where B1 is sandstone-rich, but this boundary is less easily determined in the north where the sediments are muddier. A major sandstone body 100 km long and about 35 km wide is centred on Twp. 65 Rge. 5 W6 (Figure 13). Core in 9-29-64-3W6 (Line C-C′, Figure 9) located near the seaward margin of this sandbody reveals a 4.5 m thick, bioturbated sandier-upward delta-front succession of type B; 35 km further seaward (6-27-63-1W6; Line C-C′, Figure 9), sequence B1 is represented simply by thin-bedded prodelta mudstone of facies 1. A narrow ribbon of thin sandstone is interpreted, on the basis of log character, to extend northward from the main sandstone body. Exposure at Erin Lodge reveals a type A facies succession capped by clean delta-front sandstone (Line O-O′, Figure 10). The broadly linear shape of these sandstone bodies is interpreted to indicate moderate wave influence on the delta front, although the scarcity of HCS and SCS suggest that large waves rarely influenced the coast, perhaps due to attenuation in shallow water offshore. The local thickening of sandstone to 4 m in Twp. 65 Rge. 5 may indicate the position of a major river mouth.

| Sequence B2
The marine transgressive limit of sequence B2 lies 100-150 km updip from the B1 lowstand shoreline and 20-30 km updip from the transgressive limit of B1 (Figures 13 and 16). Sandstone is restricted to the area between the transgressive and regressive shoreline limits. The regressive limit of the B2 delta front is close to that of B1 and has an overall NNE-SSW trend ( Figure 13). This palaeoshoreline trend is corroborated by palaeocurrent measurements in outcrop, where wave ripple crests are broadly parallel to shore whereas current ripples and groove casts are broadly shore-normal. Two main sandstone-rich areas in B2 lie close to the regressive limit of the delta front. These sandstone bodies are separated by a sandstone-poor area that directly overlies the main sandbody in sequence B1. It therefore seems likely that distributaries that fed the B2 deltas were diverted by the subtle topography resulting from differential compaction over the underlying sandbody.
Sequence B2 has also been mapped east of the 6th Meridian (118°W), where it forms a single, mudstone-dominated, subtly upward-coarsening succession that extends to at least the 5th Meridian (114°W). Within this part of the F I G U R E 9 Core cross-sections C2 (strike) and C3 (dip oriented), located in Figure 1 sequence, a NNE-SSW elongate sandstone body up to 7 m thick has been mapped between ranges 8 and 12 W of the 4th Meridian ( Figure 13, inset map). It is not possible to relate this eastern sandstone body to those sandstones mapped further west, in the study area of Hay (2006), and it seems likely that it represents a separate deltaic system that prograded from the north.

| Sequence B3
Sequence B3 has a 'ragged' log profile over much of the study area and neither a distinct transgressive mudstone sheet nor a well-developed regressive delta-front sandstone are recognizable. In consequence, neither the transgressive or regressive limits of the shoreline can be traced. All that can be reliably mapped is the zero-edge of the sequence in the south east, which is displaced an average of about 40 km updip relative to the regressive limit of B2 (Figures 14 and 16). In the western and central parts of the study area, sequence B3 consists of a 1-2 m thick, waxy, rubbly grey-green sideritic palaeosol with roots, organic debris and dinoturbation. Only in the east does B3 form a laterally-persistent, sandier-upward succession of type B, less than 3 m thick, suggestive of deltaic deposition in shallow water. Sandstones are patchy and generally less than 1 m thick, with a few patches up to 4 m thick in the west (Figure 14). Sequence B3 therefore appears to be composed primarily of shallow-marine to non-marine deposits with a strong pedogenic overprint.
Of the 970 well logs examined, 53 contained sharp-based and spatially isolated sandstone bodies up to 16 m thick that all 'hang' from the top surface of sequence B3 (Figure 14). Sandstones that exceeded 10 m thick were interpreted as palaeovalley fills; thinner sharp-based sandstones overlap in thickness with channel-fills and were arbitrarily excluded from the potential 'valley-fill' category. The palaeovalley fills, which average 13 m thick (range 10-16 m), could not be mapped with confidence, despite dense drilling, and hence are interpreted to be narrow, probably <1 km wide. This contrasts with the palaeovalley fills in Dunvegan allomembers H to E which average 21 m thick, typically are 1-2 km wide and are much more easily mapped (Plint, 2002). The presence of palaeovalley fills at the top of B3, and an extensive intervening palaeosol strongly suggest that F I G U R E 1 0 Approximately dip-oriented outcrop cross-section O1 on the Peace River, and O2 in the Foothills, located in Figure 1 | 539 HAY And PLInT the upper surface of that sequence is an important subaerial unconformity that appears to record a period of protracted pedogenesis and fluvial incision (McCarthy and Plint, 1998). The apparent absence of significant sand accumulations at the seaward end of these palaeovalleys might indicate that they were confined to the lower delta plain, were fed by local rainfall, and did not connect to trunk drainage systems (Wood et al., 1993).

| Sequence A1
The marine transgressive limit of A1 lies about 20 km updip (north west) of the transgressive limit of sequence B2 (Figures 14, 16 and 17). In the west and north, facies succession D dominates, providing evidence for brackish and freshwater bay and lake environments. To the south and east, sequence A1 comprises facies successions A and B, representing F I G U R E 1 1 Outcrop cross-section O3 in the vicinity of Grande Cache, located in Figure 1 F I G U R E 1 2 Stratigraphic logs summarizing the five types of facies succession recognized in Dunvegan allomembers A and B, with an interpretation of the depositional environments wave-influenced marine delta-front environments. In the north west, the transgressive limit coincides with a few large lobate sandstone bodies that may represent river-dominated highstand deltas (cf. Bhattacharya and Walker, 1991b). In the north, there is little sandstone near the progradational limit of the shoreline, which is interpreted to lie at a marked thinning of the sequence. To the south, however, sequence A1 is sandier and the progradational limit coincides with a N-S trending sandstone, 2-3 m thick that, unfortunately, is not cored (Figure 14). This sandstone apparently extends into outcrop at Copton Creek where it comprises two sandier-upward successions, the lower being heavily bioturbated whereas the upper shows wave ripples, gutter casts and HCS, suggestive of significant wave influence. Further landward, at Lynx Creek and Torrens River, sequence A1 comprises non-marine facies forming succession D (Line O2, Figure 10).

| Sequence A2
The marine transgressive limit of A2 trends NE-SW and lies 15-20 km updip from the A1 transgressive limit, and up to 100 km updip from the A1 regressive limit (Figures 15 and  16). Towards the north, sequence A2 onlaps onto A1 and does not appear in outcrop on the Peace River; the onlap limit is assumed to approximate the marine transgressive limit. Although A2 contains no sandstone bodies near the marine transgressive limit, two large, approximately linear zones of sandstone lie further seaward (Figures 15 and 16). These sandstones are fine-grained, up to10 m thick and contain wave ripples, gutter casts, planar lamination and HCS suggestive of a relatively strong wave influence; other cores show decimetre-scale cross-bedding (16-5-69-6W6, 11-3-68-6W6 in line C3, Figure 9). Outcrop sections as well as logs and cores show that these A2 sandstones are commonly sharp-based, sometimes with gutter casts at the basal contact (e.g. Lynx Creek, Copton Creek, Figure 10; 10-15-71-11W6 in line C2, Figure 9; 16-5-69-6W6 and 11-3-68-6W6 in line C3, Figure 10). The two main sandstone bodies are not separated by transgressive mudstone and hence are interpreted to be part of a single progradational package. It is inferred that the more northerly sandbody was deposited first, with continued regression (and relative sea-level fall?) leading to deposition of a second linear delta-front sandstone further to F I G U R E 1 3 Sandstone isolith maps for sequences B1 and B2 showing distribution of 'clean' sandstone (contour interval = 1 m), transgressive limits of marine mudstone (error estimate ± 5 km) and regressive limits of marine delta-front sandstone (error estimate ± 2 km). Palaeocurrent data are from equivalent units in outcrop. Delta-front sandstone bodies in B2 were not deposited directly above those in B1, suggesting accommodation was influenced by differential compaction over the buried sandbody the south east (Figures 15 and 16). Progradation of the A2 delta-front sandstone appears to have been limited by the underlying A1 sandstone (Figure 16), suggesting the existence of relict topography resulting from differential compaction.

| Sequence A3
The transgressive limit of sequence A3 lies 10-40 km updip from the transgressive limit of A2, and over 100 km updip from the regressive limit of A2 (Figures 15 and 16). The transgressive limit is recognized in outcrop on the Peace River near the Hines Creek 2 section that approximates the boundary between sandier-upward bay-fill deposits to the southeast from a palaeosol-dominated succession to the north-west (Line O1, Figure 10). No significant sandstones are associated with the marine transgressive limit. Regressive deltafront sandstone is well-developed in sequence A3, forming a linear NE-trending sandbody about 130 km long and 25 km wide ( Figure 15). The seaward margin of the A3 sandstone appears to have been constrained by two sandstone bodies in the underlying sequence A2 (outlined in Figures 15 and  16), that may have formed subtle topographic relief that localized the position of the A3 deltaic bodies. The pronounced linear shape of the A3 sandstone suggests a relatively strong wave influence. Seaward (south east) of the main sandbody, sequence A3 consists of a single, weakly developed siltierupward succession that thins to the south east. Rare, sandstone-filled palaeovalleys, up to 19 m thick, are present at the top of A3, suggesting the presence of another unconformable surface.

| EVOLVING PATTERNS OF SUBSIDENCE
West of ca 120°W, Dunvegan allomember C comprises mainly delta-plain facies and has an essentially tabular geometry, varying subtly between 5 and 10 m thick over an area of ca 200 × 250 km (see fig. 15 of Hay and Plint, 2009). Only to the east of ca 120°W does allomember C include marine deltaic facies that form a package with a F I G U R E 1 4 Sandstone isolith maps of sequences B3 and A1. Sequence B3 contains little delta-front sandstone, and instead the seaward limit of the sequence is plotted at the zero-edge of the unit. The top surface of B3 is a thick palaeosol interpreted as a subaerial unconformity that is incised by numerous narrow (<1 km) valleys (orange dots). Well density is not sufficient to allow valleys to be mapped with confidence. Transgressive marine mudstone and regressive delta-front sandstone in sequence A1 (minimum contour value = 2 m, interval = 1 m), are readily mapped, and in outcrop, gutter casts are broadly shore-perpendicular whereas wave ripple crests are shore-parallel subtly sigmoidal geometry that thickens to fill accommodation left unfilled by allomember D. In contrast to the tabular updip geometry of allomember C, the updip portions of allomembers B and A thicken westward towards the orogen ( Figure 18). Evidence from outcrop studies Plint, 2003;Lumsdon-West and Plint, 2005), indicates that the Dunvegan coastal plain was dominated by anastomosed river systems, modern examples of which are characterized by very low gradients (Makaske, 2001). There is little evidence that the alluvial systems that constructed the delta plain of allomembers C, B and A varied over time, and hence it is inferred that the alluvial gradient remained essentially similar over that time interval. Thus, the change in the isopach pattern between allomembers C, B and A is inferred to reflect primarily a change in the subsidence pattern of the proximal foredeep, rather than a steepening of the gradient, and up-building of the alluvial surface. Allomember C is inferred to have accumulated under an essentially static subsidence regime, whereas allomembers B and A show the onset of a new phase of flexural subsidence that initially formed an arcuate, westward-thickening depocentre in North East British Columbia ( fig. 16 of Hay and Plint, 2009).
The pattern of increasingly widespread subsidence continues in the immediately overlying shallow-marine deposits of the 'A-X' unit of the basal Kaskapau Formation (Figure 1). This unit forms a north-west to south-east elongate, northeast-thinning wedge that extends ca 700 km parallel to the present deformation front ( fig. 6-18 of Tyagi, 2009;fig. 38 of Plint, 2019). The dramatic, ca 90° anticlockwise rotation of isopleths from Dunvegan C to the A-X unit is interpreted to reflect the abrupt onset of subsidence, driven by thickening of the orogenic wedge as a result of north east-directed thrusting linked to transpression across a restraining bend in the Tintina-Rocky Mountain Trench strike-slip fault system (Price, 1994). This crustal shortening was ultimately related to a change in the convergence direction between the Farallon and North American plates (Plint et al., 2012).
Isopach maps of individual sequences (Figures 19 and 20), do not show the pronounced westward-thickening seen at the allomember scale, mainly because the regional pattern is masked by differential compaction of localized sandstone bodies. The sequence-scale maps do, however, reveal the gradual north and westward shift of the depocentre in response to both long-term marine transgression and flexural subsidence. Note that it is not possible to reliably correlate individual sequences much to the F I G U R E 1 5 Sandstone isolith maps of sequences A2 and A3 (minimum contour = 2 m; interval = 1 m). Note how the thicker sandbodies have a mutually evasive distribution with respect to immediately underlying sandstone bodies. To the north west, marine mudstone in sequence A2 grades laterally into coastal plain facies, whereas to the north east the sequence pinches out through onlap onto A1 | 543 HAY And PLInT F I G U R E 1 6 Interpreted palaeogeography of each sequence at maximum regression, based on sandstone isolith maps constructed mainly from drilling data, and including the transgressive limit of the marine shoreline interpreted from log signatures. The south west side of each block is based on facies relationships seen in outcrop along the Foothills whereas the north east side is based on exposure on the Peace River. The position of the lowstand shoreline in B3 can not be mapped because sandy lowstand deposits are absent, but it is likely to have lain close to, or seaward of the zero isopach limit because palaeovalleys are widely incised into the top of B3 (see Figure 14) and a few are incised into B2 suggesting complete emergence of B3 at lowstand. Successive deltafront sandstone bodies tend to have a mutually evasive relationship with thick sandstone in the immediately underlying sequence, suggesting that deltas tended to fill space generated by differential compaction of underlying mudstone west of 120°W (the Alberta-British Columbia border), so the full extent of thickening into the foredeep is not evident.

| SEA-LEVEL CHANGES
Although autocyclic delta switching is likely to have been responsible for some of the locally developed, metre-scale upward-shoaling successions in allomembers B and A, three lines of evidence suggest that sequences B1 to A3 were primarily of allogenic origin. Firstly, the regional extent of each of the marine transgressive mudstone blankets at the base of each sequence is of much greater extent than any of the deltaic lobes revealed by isolith maps of delta-front sandstones (cf. Bhattacharya, 1989;Bhattacharya and Walker, 1991b). Relative sea-level rise is therefore difficult to attribute to localized subsidence of abandoned delta lobes. Secondly, most of the wave-dominated delta-front sandstone bodies (facies 10) rest sharply and erosively on underlying more offshore deposits (facies 7 and 8). This relationship implies an abrupt increase in wave energy at the toe of the delta front, that may most readily be explained in terms of relative sea-level fall that led to increased wave-scour of the prodelta area, linked, possibly, to accelerated progradation ('forced regression') of the delta front (Plint, 1991). Thirdly, the presence of palaeovalleys incised into the top surfaces of sequences B3 and A3, coupled with a widespread palaeosol on the equivalent interfluve surface suggests a regional, allogenic control, possibly relative sea-level fall and/or a change in the hydraulic regime of the rivers). Relative sea-level fall can be inferred in all six sequences, based on the presence of palaeovalleys and sharp-based delta-front sandstones. The restriction of palaeovalleys to the top of sequences B3 and A3 suggests, however, that sealevel falls of greater magnitude, or greater duration, may have terminated these two sequences. Alternatively, valley incision may have been due to a change in the ratio of discharge to sediment load of rivers, as predicted by numerical models Zhang et al., 2019), recognized in Quaternary river systems (Gibling et al., 2011;Blum et al., 2013), and inferred for Cretaceous valley systems in Dunvegan allomembers H to E (Plint and Wadsworth, 2003), and in Upper Albian coastal plain strata of the Paddy alloformation (Plint et al., 2018).
Dunvegan sequence B1 shows a ca 210 km backstep from the maximum regressive limit of allomember C. Following this initial, major flooding event, sequences B1 through A3 each record transgression of the shoreline by about 80 km ( Figure 17). The transgressive limit of successive sequences is progressively offset landward by an F I G U R E 1 7 Chronostratigraphic plot and averaged shoreline migration curve, based on palaeogeographic relationships in Figures 13 through 15, supplemented by data in Hay and Plint (2009) and Bhattacharya (1989). Vertical axis is time; collectively allomembers A and B are estimated to span about 250 kyr (see Figure 21 and discussion in text). Continued updip aggradation of the alluvial portions of some sequences reflects differential updip flexural subsidence of the depocentre. The lowstand shoreline at the top of B3 must have lain at least as far seaward as the most seaward palaeovalley, rather than at the seaward margin of delta-front sandstone (see Figure 14) average of ca 15 km. Although the pattern of regressive shorelines tends to be complicated by deltaic headlands and embayments, viewed as a whole, the regressive limits of the delta front also show a progressive landward offset.
Excluding the anomalously large basal transgression of B1, each sequence shows an average shoreline excursion of about 80 km, between the maximum transgressive and maximum regressive limits.

F I G U R E 1 8 Isopach maps (contour interval = 5 m) for allomembers A and B, showing differential subsidence adjacent to the deformed belt.
At the scale of allomembers, thickening over localized delta-front sandstone bodies tends to be unrecognizable because of compensation stacking between the component sequences F I G U R E 1 9 Isopach maps for sequences B1-B3. Thicker regions tend to correspond to delta-front sandstone bodies 546 | HAY And PLInT

| Eustatic or tectonic control on sea level?
Isopach maps (Figure 18) show that the foreland basin was experiencing flexural subsidence during deposition of allomembers B and A, with the strike of the flexed surface trending approximately N-S to NNW-SSE, dipping to the west or west south west. Maps of the regressive shoreline of successive sequences (Figures 13 through 15) are interpreted to represent a broadly arcuate coastline, open to the south east. The trend of the principal delta-front sandstone bodies changes from ca N-S in sequences B1 and B2, rotating to NE-SW in A2 and A3.
If transgressions and regressions had been driven primarily by changes in the rate of flexural subsidence (or uplift), the shoreline would be expected to have migrated perpendicular to the strike of the flexed surface, that is, in an east-west sense. Mapping shows, however, that the shoreline was oriented at an oblique angle to the strike of the flexed surface, approaching perpendicular in the case of sequence A3. The palaeogeographic maps show that, despite subsidence being down to the west or west southwest, the sedimentary system maintained a depositional surface that sloped down to the east or south-east. This geometric relationship implies that the rate of sediment supply was adequate to fill all the updip accommodation generated by tectonic subsidence, and that sediment dispersal (probably dominated by storm-driven processes), was able to fill accommodation as fast as it was generated, even in the most proximal part of the foredeep. These observations suggest, therefore, that eustasy exerted the principal control on shoreline movement, alternately driving the shoreline back and forth across a depositional surface consistently inclined to the east or south-east.

| Amplitude of sea-level change
If the slope of the delta plain can be estimated, it is possible to make a first-order estimate of the magnitude of sea-level change for each sequence. The dominantly anastomosed character of the rivers that constructed the delta plain during transgressive and highstand time suggests a gradient ranging from 30 to as little as 2 cm/km (Makaske, 2001). For comparison, Lin and Bhattacharya (2017) suggested, on the basis of empirical relationships, that Dunvegan trunk rivers had a gradient of the order of 5 cm/km. Using the observed average transgressive distance of ca 80 km per sequence (maximum regressive to maximum transgressive shoreline; Figure 17), and postulating a range of delta-plain gradients, from 30, 20,10 and 5 cm/km, it is possible to calculate relative sea-level rise of ca 24, 16, 8 and 4 m, respectively. Given that each sequence is of the order of 10 m thick and is capped by a subaerial surface, it would appear that a gradient of 5 cm/km would provide too little accommodation to account for the observed thickness of rock, even allowing for modest hydro-eustatic subsidence. Conversely, a gradient of 30 cm/km apparently provides too much accommodation. The available data therefore suggest that each sequence developed primarily in response to a eustatic rise in the range 8-16 m, flooding a coastal plain with a gradient in the range of 10-20 cm/km. The fact that most delta-front sandbodies are sharp-based suggests that progradation of the more seaward parts of each sequence (i.e. during the falling stage systems tract), was promoted by relative F I G U R E 2 0 Isopach maps for sequences A1-A3, where thicker regions correspond to thicker sandstone bodies. Collectively the isopach maps of sequences B1-B3 and A1-A3 show the progressive displacement of the depocentre to the north and west in response to long-term eustatic rise, coupled with flexural subsidence sea-level fall that resulted in wave-scour of the prodelta region and forced regression of the delta front.
Studies of other middle and late Cenomanian shallow and marginal-marine clastic sequences, both in Europe (Uličný and Špičáková, 1996), and North America (Uličný, 1999;Laurin and Sageman, 2007;Laurin et al., 2019), document depositional sequences consistently in the range 4-12 m thick, inferred to have been generated by high-frequency sealevel cycles in the range 10-20 m.

| Timescale of sea-level change
Ammonites are very rare in the Dunvegan Formation, presumably because of unfavourable ecological conditions, and hence relating Dunvegan allomembers to the standard Western Interior ammonite biozones ( Figure 21) is not possible. Foraminifera provide only a very low-resolution biostratigraphic control. Only two reliable tie points are available: The X bentonite, identified by Tyagi et al. (2007) and dated by Barker et al. (2011) to 95.87 ± 0.1 Ma lies near the middle of Dunvegan allomember C, occupying a position near the top of the Acanthoceras amphibolum Zone. The first occurrence of Dunveganoceras albertense is at the base of Doe Creek allomember 8 (Plint and Kreitner, 2019), with an interpolated age of 95.01 Ma (Ogg and Hinnov, 2012). Between these two tie points, separated by ca 860 kyr, lie 20-21 simple depositional sequences (upper part of Dunvegan allomember C, allomembers B1-B3, and A1-A3, and, in the overlying Kaskapau alloformation, allomembers A-X1 to A-X5, and Doe Creek allomembers 1, 1A, 2, 2A, 3, 4, 5, 6 and 7), suggesting that each sequence spans an average of about 41 kyr ( Figure 21).
For comparison, the six slightly younger sequences 3A, B, 4, 5 and 6A and B mapped in the middle Dakota Formation, were each interpreted to span from 100 to 200 kyr (Uličný, 1999). Recent radiometric dating (Barclay et al., 2015) suggests that the same six sequences have a duration closer to 100 kyr. The overlying 'genetic sequences' S2 and S3 in the upper Dakota Formation (Laurin et al., 2019) each span about 100 kyr (Figure 21), but are composed of higher-frequency sequences that have a periodicity of about 20 kyr. Thus, although it is difficult to establish precise age control for the sequences in Dunvegan allomembers B and A, it seems very likely that they developed on a timescale of <100 kyr.  Ogg and Hinnov (2012) and Laurin et al. (2019). Middle column shows the succession of allomembers in North West Alberta and adjacent British Columbia (compiled from Varban and Plint, 2005;Kreitner and Plint, 2006;Tyagi et al., 2007;Hay and Plint, 2009;Barker et al., 2011;Van Helmond et al., 2016;Plint, 2019). Orange lettering denotes reliable age and biostratigraphic control points for the Canadian succession. Right column shows contemporaneous depositional sequences interpreted in the Dakota Formation, Utah (after Uličný, 1999;Laurin et al., 2019). The age range of Uličný's sequences 2 through 6B have been recalibrated using the radiometric dates in Barclay et al. (2015) 548 | HAY And PLInT delta-plain lakes, palaeosols and rare palaeovalley fills. Sandy delta-front and mouth bar facies contain little HCS or SCS and tend to be dominated by current and combined-flow ripples, suggesting that wave energy was relatively low. Prodelta and delta-front facies may have low or high levels of bioturbation that probably reflect varying degrees of infaunal suppression by reduced salinity, high turbidity or high sedimentation rate, linked to proximity of distributary mouths. 3. Isolith maps show that the volume of sandstone within successive sequences is highly variable. Delta-front sandstone bodies show a range of planform morphologies, ranging from localized and lobate (a few to a few 10s of kilometres wide) to highly elongate (up to 150 km long). More localized sandstones, prevalent in lower sequences, are interpreted to reflect deposition around major river mouths and subject to modest lateral reworking of sand, whereas elongate sandbodies, that are best developed in the upper two sequences, may reflect more open-marine conditions and more effective wave reworking as transgression proceeded. 4. The top of sequence B3 is incised by narrow (probably <1 km), sandstone-filled palaeovalleys up to 16 m deep.

| CONCLUSIONS
Intervening wells show a thick muddy palaeosol, interpreted as an interfluve surface. The top of sequence A3 is also incised by rare, sandstone-filled valleys up to 19 m deep. Valley incision may have been a response to relative sea-level fall and/or a change in the ratio of discharge to sediment load in rivers; climate change may have been a driver for both mechanisms. 5. The transgressive and regressive limits of most sequences can be mapped from wireline log signatures, and at outcrop. The shoreline described a broad arc, open to the south-east, with the main sense of transgression and regression oriented north west-south east. Sequence B1 shows a 210 km transgressive backstep from the regressive limit of underlying allomember C2. After that initial, major transgression, sequences B2 through A3 show a fairly consistent pattern, in which the shoreline migrated landward and then seaward by an average of about 80 km through each transgressive-regressive cycle. The landward limit of each transgression is consistently offset landward by about 15 km per sequence, reflecting long-term relative sea-level rise that ultimately drowned the entire delta complex. 6. Isopach maps of allomembers B and A show thickening towards the west and south-west, implying syn-depositional flexural subsidence driven by the tectonic load of the Cordillera. 7. Although tectonic subsidence resulted in flexure down to the south-west, the sense of shoreline movement (north west-south east) was near-perpendicular to that 'tectonic slope'. This implies that, despite flexural subsidence, the depositional system maintained a consistent bathymetric slope to the south-east, maintained by efficient sediment redistribution processes. The sediment supply was adequate to fill all tectonically generated accommodation, even in the most proximal part of the foredeep. On the basis of these geometrical relationships, it is concluded that the six studied sequences were generated primarily by eustatic change that drove the shoreline back and forth across a surface inclined to the south-east. Most deltafront sandstone bodies are sharp-based, implying that coastal progradation was driven, at least in part, by relative sea-level fall. 8. The delta plain was characterized by anastomosed rivers that typify very low-gradient alluvial systems. Simple calculations suggest that eustatic excursions in the range 8-16 m across a terrestrial surface inclined at 10-20 cm/ km would have been adequate to generate each of the six sequences, each of which has an average thickness of about 10 m. 9. Although geochronologic and biostratigraphic data are sparse for the Cenomanian strata in the study area, available evidence suggests that the six simple sequences forming Dunvegan allomembers B and A may each have had an average duration of ca 41 kyr. This cycle frequency is comparable to that observed in other Late Cenomanian clastic successions in the USA and Europe, and implies climate modulation at the high-frequency end of the Milankovitch band.