Abstract
At least four horizons of enhanced anoxia (anoxic horizons; AHs) are recognized in the uppermost Eifelian–Middle Frasnian mudrock-dominated strata of the Mackenzie Valley and Peel area of NW Canada. Aluminium-normalized Mo and U logs in two cored sections reveal AH-I at the Eifelian–Givetian boundary, AH-II in basal Frasnian, and AH-III and AH-IV bundled in the Middle Frasnian interval. These four horizons are characterized by attenuated siliciclastic components. Spectral gamma-ray K + Th and U are the best tools to trace these horizons in wells and outcrops. AHs are biostratigraphically correlated with ‘black-shale events’ in several basins of the world. Depositional environment is depicted as a stratified basin where the water-column chemocline defined co-sedimentation of anoxic mudrocks in topographic lows and oxic grey shales and carbonate banks on seafloor elevations. Based on inductively coupled plasma elemental data from 1687 samples, siliciclastic-lean basinal mudrock units that host AHs are strongly enriched in Mo (median EFMo ∼ 97–172 EFMo/EFU ≈ (3–3.5) × SW, where EFMo and EFU are respectively Al-normalized Mo and U in enrichment factor notation and SW is average present-day seawater value) compared with siliciclastic-rich units (median EFMo ∼ 17–37) and show strong EFU/EFMo covariation (r ≈ 0.8 in Canol Formation and Bluefish Member). Supported by a lack of geological evidence for an oceanographic barrier, this enrichment indicates unrestricted water exchange with Panthalassa. At the same time, development of oligotrophy is indicated by a lack of P enrichment and weak to non-existent enrichment in Zn and Cu. These features are reconciled through a model by earlier workers that involves a global shift to a warm greenhouse mode with slowed oceanic convection, expanded oxygen minimum zones and a failure of nutrient resupply from the upwelling. The onset of mass degassing in continental large igneous provinces represents a potential trigger for this mid-Devonian shift. Devonian black-shale events in this scenario represent genuine oceanic anoxic events marking hothouse episodes in their nascent form.
Supplementary material: Details of methods, analytical protocols and data scatterplots, stratigraphic cross-sections showing traceability of anoxic horizons, and inductively coupled plasma elemental and Rock-Eval 6 data used in this study are available at: https://doi.org/10.6084/m9.figshare.c.4212428
The concept of oceanic anoxic events (OAEs) links simultaneous, geologically brief (a few hundred thousand years, rarely more protracted) episodes of black-shale deposition in Mesozoic–Cenozoic shelfal basins with oceanic records obtained originally through the Deep Sea Drilling Project (Schlanger & Jenkyns 1976; Jenkyns 2010). Mechanisms leading to OAEs are seen in reciprocally linked processes: (1) oxygen minimum zones (OMZs) under oceanic areas of increased productivity expand with global warming, and euxinia expands in OMZ centres (Gilly et al. 2013); (2) the thermal component of thermohaline circulation weakens during greenhouse periods and probably shuts down episodically in extreme hothouse conditions triggered by massive CO2 release in large igneous provinces (LIPs) (Bond & Wignall 2014); (3) ocean stratification and thermal expansion (Haline Euxinic Acidic Thermal Transgressions or HEATT; Kidder & Worsley 2010) capable of raising global sea-level by a maximum of c. 20 m; (4) nutrient cycle imbalances with consumable P, N, Fe, S and Mo trapped and partly precipitated in expanded anoxic waters (Kidder & Worsley 2010).
Paleozoic OAEs are harder to identify as virtually no oceanic-crust record is available and geological-time resolution available for stratigraphic packages is usually coarser than needed. The best documented Paleozoic OAE is associated with the two-step latest Permian extinction and delayed Early Triassic recovery, acknowledged as the worst mass extinction of the Phanerozoic (Wignall & Twitchett 2002; Kidder & Worsley 2004, 2010; Meyer et al. 2011; Grasby et al. 2016; Bond & Grasby 2017). It has been recognized for decades that expansion of anoxic facies in Devonian shelfal strata may have a causal link with perturbations in ecosystems expressed in most severe form in mass extinctions (Taghanic and Frasnian–Famennian extinctions in Fig. 1), which became imprinted in the chart of joint biotic–abiotic Devonian events (House 1983, 1985, 2002; Walliser 1996; Sandberg et al. 2002; Bond et al. 2004; Becker et al. 2016). However, black-shale events in the Devonian are more numerous than mass extinctions, and the severity and duration of these anoxic events do not appear to correlate with the severity of biotic extinctions (Fig. 1).
Occurrence of anoxic black shales in six global regions plotted against sea-level curve, δ13C curve from skeletal carbonate (both compiled by Becker et al. 2016), low-latitude sea-surface temperatures (LLSST) from δ18O of conodont apatite (Joachimski et al. 2009), extinction–origination curves for marine fauna (Paleobiology Database; Melott & Bambach 2014), and age ranges of large igneous provinces (LIPs; Kravchinsky 2012; Wu et al. 2013; Ricci et al. 2013). Calibration of radiometric ages from the Geological Time Scale 2012 (Becker et al. 2012). For LLSST, bold centre-line is the locfit regression and fine lines give the 95% confidence interval. Alternative conodont zones in Givetian–Frasnian (Narkiewicz & Bultynck 2007; Becker et al. 2016): tim., timorensis; rh.-v., rhenanus–varcus; ans., ansatus; lat., latifossatus/semialternans; nor., norrisi; pr., pristina; rot., rotundiloba; rug., rugosa. Depophase indexing in the sea-level curve is retained from Johnson et al. (1985). Sources for chronostratigraphic position of anoxic shales: NW Canada: references in this study (L.Vermill. is the lower Vermillion Creek Member, M.H.I. is the middle Hare Indian in the Prohibition Creek Member; *black-shale units with poorly constrained age); Western North America: Wendte & Uyeno 2005; Dong 2016; Knapp et al. 2017; Dong et al. 2018; Eastern–Central North America: Ver Straeten et al. 2011; Brocke et al. 2015; Enomoto et al. 2016; Klapper & Kirchgasser 2016; Central and Western Europe: House 1996; Bond et al. 2004; Buggisch & Mann 2004; Buggisch & Joachimski 2006; Yans et al. 2007; Brocke et al. 2015; Kaiser et al. 2015; Becker et al. 2016; Eastern Baltica–Western Urals: House et al. 2000; Abramova & Artyushkova 2004; Manner & Shuvalova 2007; Artyuszkova & Maslov 2008; Gatovskii et al. 2016. NW Africa: Lüning et al. 2004; Kaiser et al. 2007, 2015; Aboussalam et al. 2015; Mahboubi & Gatovsky 2015. Block length in LIPs denote oldest and youngest absolute ages, and width reflects relative volume of emplaced basalts. F-F, Frasnian–Famennian.
Black shales are particularly widespread in Givetian–Tournaisian shelfal basins with corresponding significance for petroleum source rocks. Late Devonian–Tournaisian source rocks account for over 8% of the world's petroleum reserves (Klemme & Ulmishek 1991; Trabucho-Alexandre et al. 2012). Some black-shale basins span the entire Givetian–Famennian time interval of c. 30 myr (Fig. 1), suggesting that protracted hypoxic to anoxic conditions outlived recognized ‘anoxic events’. This is somehow linked to the co-deposition of basinal laminated sediments and isolated carbonate banks (reefs, pinnacles, mud mounds) that were able to keep their growth close to sea-level. This co-deposition appears to be very common in Givetian–Tournaisian shelfal basins (Playford 1980; Playford et al. 2009; Corlett & Jones 2011; Gatovskii et al. 2016; Knapp et al. 2017), including the study area (Kabanov & Gouwy 2017).
In the ‘Devonian eustatic sea-level curve’ of Johnson et al. (1985) and its later modifications (e.g. Sandberg et al. 2002), anoxic horizons are seen as a result of genuine deepening or global transgressions (Arthur & Sageman 2005; Bond & Wignall 2008). Interpretations of major cycles in Devonian basinal strata of the study area and adjacent regions follow this concept by default (e.g. Fraser & Hutchison 2017; Dong et al. 2018). Some workers have referred Devonian anoxic horizons to OAEs (Algeo et al. 1995), but this idea remains unpopular owing to a lack of sufficient evidence for the nature of these ‘black-shale events’ (Bond & Grasby 2017). The known ice-free character of pre Late Famennian global climate implies that ‘major cycles’ of the Devonian are non-glacioeustatic (e.g. may be a eustatic response to growth and decay of mid-ocean ridge systems; Johnson et al. 1985). On the other hand, the fluctuating redox state of the World Ocean seems alone to be able to govern simultaneous shifts to anoxic facies (Meyer & Kump 2008), which is discussed below involving novel geological observations, conodont age and elemental geochemistry data collected from the uppermost Eifelian–Frasnian black shales of northwestern Canada.
Geological context
The study area is situated in the northwestern part of the ancestral North America (Fig. 2a and b). The ancestral North America is a reference to the main part of the Laurentian continental block that remained relatively stable throughout the Phanerozoic (Gabrielse & Yorath 1991), and it is firmly embedded in the lexicon of North American geology (Colpron & Nelson 2011; Colpron et al. 2006, 2007; Nelson et al. 2013). The Paleozoic sedimentary wedge of the northwestern ancestral North America bordered on the west by the Richardson and Selwyn basins is referred to as the Mackenzie Shelf (Fig. 2; Gabrielse & Yorath 1991; Cecile & Norford 1993). It consists of the Early Paleozoic–Middle Devonian (Eifelian) platform dominated by shallow-water carbonates, the Horn River Group, and the foreland basin fill of the Imperial and Tuttle formations (Fig. 2b; Hadlari et al. 2009; Kabanov & Gouwy 2017). Detrital apatite-fission track and apatite–zircon (U–Th)/He data from the Mackenzie Plain and its vicinity indicate that subsidence continued into the Late Paleozoic and Triassic when the Imperial Formation was buried at depths corresponding to 120–160°C heating, and significant unroofing occurred in the later Mesozoic prior to Laramide orogeny and associated clastic load (Issler et al. 2005; Powell 2017).
The study area (oval) in global palaeogeographical and tectonic context. (a) Palaeogeography and plate-tectonic reconstruction at 380 Ma (clipped Mollweide projection), modified from DeepTimeMap collection supplied by ©2016 Colorado Plateau Geosystems Inc. Blank cross-hatched area occludes Alaska, the position of which is debatable (Lane et al. 2016). Other global regions of black-shale deposition in reference to Figure 1: (WB) Williston Basin; (HB) Horn River–Liard Basin; (AB) Appalachian foreland basin; (PB) Permian Basin with Woodford Shale; (TP) Timan–Pechora and (VU) Volga–Urals are contiguous oil and gas provinces of Eastern Baltica–Western Urals; (NA) Northwestern Africa; (WCE) Western–Central Europe. LIPs are indexed by red: DDRB is Dnieper–Donets Rifted Basin. The hatched eastern part of the Mackenzie shelf, now part of the Canadian Shield, was covered with thick Devonian strata (Ault et al. 2013). Tr. C. Arch, Transcontinental Arch of North America. (b) Map of cordilleran terranes (modified from Colpron & Nelson 2011). Basinal and platformal areas of the ancestral North America are adapted to Devonian palaeogeography; only terranes discussed in this paper are labelled. Orange dashed lines are deformation fronts of the Devonian Romanzov (R) and Ellesmerian (E) orogenies (Lane 2007).
Palaeotectonic setting of Horn River Group deposition
The Early Paleozoic–Eifelian carbonate platform of the Mackenzie Shelf, culminating in deposition of the Hume Formation, is characterized as a passive-margin stratal succession deposited following Neoproterozoic–Early Cambrian rifting (MacLean 2011). The Early Devonian–Eifelian carbonate shelf sloped along its western margin into the Selwyn shale basin and the Richardson Trough (Fig. 2b; Morrow 2012). Deposition of conformable strata in a tectonically quiet setting continued into the Givetian, although carbonate sedimentation switched across the Mackenzie Shelf into the basinal organic-rich mudrocks of the Horn River Group. The Hume carbonate platform and the overlying anoxic shales of the Bluefish Member are traced as far north as central Banks Island of the southwestern Arctic Archipelago (Kabanov 2018). The Headless Member at the base of the Hume Formation (Fig. 3b) records transgression coincident with the acme of the Romanzov Orogeny, which is traced in the northern Yukon and adjacent Alaska; however, no deformation reached the study area (Fig. 2b; Lane 2007; Beranek et al. 2010).
(a) Geographical spread of the Horn River Group (on NWT side) and Canol Formation (on Yukon side) between 64 and 68°N; crosses are outcrops (1, Trail River (TR); 2, Rumbly Creek; 3, Powell Creek); diamonds are wells (O06 is Loon Creek O-06; N09 is Little Bear N-09). Palaeogeographical areas: WOB, western off-bank area; BAT, bank-and-trough area; SOB, southern off-bank area. (b) Cross-section (black line in (a)); top of Hume Formation as datum; wells from west to east: Cranswick YT A-42, Cranswick A-22, S. Ramparts I-77, N. Ramparts A-59, Ramparts River F-46, Hume River I-66, Hume River D-53, Carcajou L-24, Maida Creek F-57, Hoosier F-27, NWB is Norman Wells oilfield, Little Bear N-09, Bluefish A-49, and Bracket Lake C-21. Stratigraphic members in SOB: (fc) Francis Creek, (pc) Prohibition Creek, (vc) Vermillion Creek; (dc) Dodo Canyon, (ml) Mirror Lake, and (lc) Loon Creek. NRTF is Norman Range thrust fault (other tectonic elements are not shown). The Canol Formation in WOB area is provisionally subdivided into (1) lower, (2) middle and (3) upper traceable units or information members (cross-section B–B' in supplementary material).
The turbidite system of the Ellesmerian foreland basin, the depositional setting of the Imperial Formation (Hadlari et al. 2009), encroached into the study area in the late Middle–early Late Frasnian, but associated deformation did not reach present-day 67°N (Fig. 2b). Syndepositional folding in the Imperial Formation is known only along the eastern margin of the Mackenzie Delta (Lane 2007).
Lithostratigraphy of Horn River Group
The latest Eifelian–Frasnian Horn River Group occurs across a vast area in the Northwest Territories, in both the cordillera and the adjacent interior plains (Morrow 2012), and is equivalent to the Canol Formation of Northern Yukon (Fig. 3). In the study area, the occurrence of thick (150–400 m) Hare Indian and Ramparts strata defines the bank-and-trough palaeogeographical area, or BAT (Fig. 3). This succession of fossiliferous grey shales, siltstones and limestones conformably overlies the basal black-shale unit of the Hare Indian Formation called the Bluefish Member. The latter onlaps benthic fossiliferous limestones of the Hume Formation with a drowning unconformity sensu Schlager (1989). The Bell Creek Member of the BAT area is a thick (up to 190 m) succession of grey, variously calcareous shales and siltstones representing clinoform basin fills or ‘deltaic shale lobes’ (Muir & Dixon 1985) sourced from the easterly located Laurentian landmass.
The Ramparts Formation (up to 163 m thick) is composed of three main parts and is best described as carbonate banks outgrowing thicker deltaic lobes of the Hare Indian Formation. The lower platform (or lower ramp) member is an argillaceous and silty fossiliferous limestone intergrading with, and overlying, the upper Bell Creek Member of the Hare Indian Formation. The middle Carcajou unit is a thin (<3.0 m) limestone containing benthic fossils and is distinguished from underlying and overlying Ramparts limestones by its dark argillaceous and bituminous matrix. The Carcajou is identified in many wells by its elevated gamma-ray response. In some sections at Norman Wells, the limestone of the lower platform member is not sufficiently developed, and the Carcajou marker overlies the calcareous fossiliferous siltstone of the Hare Indian Formation. The upper Kee Scarp Member (also known as the reef member; Pugh 1983) is defined as a portion of the Ramparts developed above the Carcajou marker. This member forms spatially restricted carbonate banks and pinnacles and is composed of relatively clean limestones with stromatoporoids and corals. In the best studied examples, it reveals the backstepped, tapering-upward stacking pattern of subtidal–intertidal cycles or parasequences (Muir et al. 1985; Yose et al. 2001). Minor lithostratigraphic units historically appended to the Ramparts Formation include the Charrue sandstone and the informal ‘allochthonous limestone member’ (Pugh 1983). The latter refers to thin (up to 5.5 m) strata of local occurrence composed of graded laminated calcisiltites and bioclastic calcarenites partly interbedded with Canol-type black shales. These allochthonous limestones were interpreted as off-reef debris (MacKenzie 1970, 1973; Pugh 1983).
In the central Mackenzie Valley, south of Norman Wells, the Hare Indian–Ramparts section thins into a black-shale dominated package, 20–50 m in thickness. Kabanov & Gouwy (2017) referred to this basinal area between Norman Wells and the axis of the Keele tectonic zone as the southern off-bank area or SOB. Immediately west of the 130° meridian in the study area, the thick Hare Indian–Ramparts succession of the BAT grades into thin black shales similar to those of the SOB (Fig. 3). Further west, across 132° meridian, the entire Hare Indian Formation is considered to be merging into the Canol Formation as its gamma-ray and resistivity well log signatures disappear in response to westward attenuation of siliciclastic components (western off-bank area or WOB in Fig. 3).
The Canol Formation is composed of laminated siliceous pyritic shales with minor siltstone, calcareous and dolomitic lithologies. The Canol Formation is 60–120 m thick in off-bank areas, but thins to a few metres above the Ramparts carbonate banks (Fig. 3) and disappears on the tops of the tallest carbonate banks intersected by wells (Pugh 1983). Contact of the Canol and Imperial formations is gradational, and its correlation across the study area remained unsettled for decades. Recent refinements for the SOB area are reflected in Figure 3b (Kabanov & Gouwy 2017).
Database and methods
One hundred and seventy-seven wells used in this stratigraphic study represent nearly the complete inventory of post-1940s exploration wells that penetrate the Horn River Group between 64° and 68°N. Several wells from the Camp Canol operation of 1942–1945 were also utilized based on the quality of their resistivity logs and their historical research value. These wartime wells, although drilled before the advent of gamma-ray and sonic logging, laid the foundation for subsurface stratigraphy (Tassonyi 1969). This combined pool of wells is augmented by measured outcrop sections surveyed with the spectral gamma-ray tool (SGR). Four sections with SGR logs and geochemical data are selected for discussion in this paper. The Trail River section of the NE Yukon (Fig. 3) is the westernmost of available sections with representative chemostratigraphic logs (Fraser & Hutchison 2017) and therefore it is included in discussion, although its correlation with the study area is complicated owing to a lack of hard age constraints and problems in traceability of geochemical and SGR signals.
Two continuously cored sections from the SOB area were sampled for inductively coupled plasma (ICP) elemental geochemistry and Rock-Eval 6 pyrolysis at 0.5 m intervals: 326 samples from Little Bear N-09 and 380 samples from Loon Creek O-06 (Table 1 and Fig. 4; see supplementary material for analytical protocols). The Trail River outcrop contains 148 samples processed using the same protocols (Fraser & Hutchison 2017). A multisource database contains 1751 ICP elemental measurements inside and adjacent to the study area, between 63.6° and 66.5°N (supplementary material). The mudrock subset (excluding the Hume and Ramparts formations) counts 1687 samples (supplementary material). The degree of pyritization proxy (DOPT) is the ratio of FeS (iron calculated from total S) to total Fe. The FeS parameter is based on the assumption that sulphur is entirely bound in iron disulphide represented by pyrite (Algeo & Maynard 2004) and possibly marcasite (Schieber 2011). The enrichment factor of a trace metal X is adopted from Algeo & Tribovillard (2009) as EF(element X) = (X/Al)sample/(X/Al)PAAS where PAAS is the average post-Archean Australian shale value of Taylor & McClennan (1985). The enrichment factor for phosphorus (EFP) is calculated in the same way. Likewise, the reference average shale value for Cd is 0.8 ppm (Sweere et al. 2016). It should be acknowledged that the comparison of enrichment factor-based proxies with oceanographic parameters is always notional as rock EFs are affected by the choice of reference value (either ‘average shale values’ or regional sourceland background). Additionally, normalization to Al, although certainly reducing detrital signal in sediments enriched in terrigenous fine material, has a range of drawbacks including possible presence of authigenic clay minerals, or spurious, often inverted, normalized values in rocks with low and highly variable Al content (Tribovillard et al. 2006). For the latter reason, EFs of trace metals are not analysed in Hume and Ramparts limestones.
Anoxic horizons (I–IV) revealed by lithogeochemical proxies in Little Bear N-09 and Loon Creek O-06 wells and their SGR correlation in Powell Creek and Rumbly Creek outcrop sections. Lithogeochemical logs are smoothed by LOWESS regression with α-tension 4% (see supplementary material for technique and data scatterplots).
Lithogeochemical metrics of lithostratigraphic units in SOB area (only proxies that are discussed in the text) based on Loon Creek O-06 and Little Bear N-09 data
The terrigenous input proxy (TIP) is the summed wt% of Al2O3, Fe2O3, K2O and TiO2; SiO2/Zr is used in regional studies to characterize biogenic silica input (Hildred et al. 2011; Pyle & Gal 2016). In SGR logs from boreholes and outcrops, the trace-metal proxies are compared against uranium signals (U(gAPI) and Un), and TIP against uranium-stripped or computed K–Th gamma (CGR(gAPI)). The latter is widely used as a proxy for siliciclastic input (Ellis & Singer 2007). The Un is the potassium-normalized U signal (Un = 0.29 × U(ppm)/K(%)) mimicking EFU (EFU is Al-normalized U in enrichment factor notation) based on strong linear covariation of Al and K in the shale subset of ICP data. More detail is given in the supplementary material.
Age brackets for the studied strata are provided by conodonts, as attempts to extract datable materials from bentonites have been hitherto unsuccessful. Conodont biostratigraphic information has been summarized by Gouwy in the study by Kabanov & Gouwy (2017) and post-publication updates have been delivered through Geological Survey of Canada (GSC) palaeontological reports (Gouwy 2016a, b, 2017a, b, 2018; see supplementary material for access information). Other recent summaries of conodont age have been given by Pyle & Gal (2016) and partly (for the base of the Horn River Group) by Uyeno et al. (2017).
Anoxic horizons
The intervals cored in the Loon Creek O-06 and Little Bear N-09 wells of the SOB area are correlated in remarkable detail using elemental proxies (Fig. 4). The Prohibition Creek–Dodo Canyon interval comprises a thick package of hypoxic–anoxic black shale where Fe and S are almost entirely bound in sulphide (DOPT ≈ 1). This main anoxic package has median total organic carbon (TOC) content around 5% (Table 1) and includes the Prohibition Creek Member and the Canol Formation (Fig. 4). It is sandwiched between the siliciclastic-rich shales of the upper Bluefish–Francis Creek members and the grey-shale clinoform of the Mirror Lake Member. The latter is characterized by strongly suppressed DOPT, EFU and EFMo, where EFU and EFMo are respectively Al-normalized U and Mo in enrichment factor notation (Table 1 and Fig. 4).
Chemostratigraphic subdivision is further refined by bundled responses of EFMo and EFU (Fig. 4). Molybdenum shows the strongest enrichment (Table 1). The EFMo–EFU spike in the middle of the Bluefish Member divides basal siliciclastic-lean and upper siliciclastic-rich parts of this unit. This spike represents the anoxic horizon I (AH-I).
The Canol Formation of the SOB is subdivided into the Vermillion Creek and the Dodo Canyon members (Kabanov & Gouwy 2017). The Vermillion Creek Member contains anoxic horizon II (AH-II) 10–12 m above its base (Fig. 4). Through U(gAPI), Un and, to a lesser degree, total gamma-ray, AH-I and AH-II are traced in many sections in the subsurface and outcrops (see supplementary material), including the Rumbly Creek outcrop of the WOB area (Fig. 4).
The Dodo Canyon Member is the unit most depleted in siliciclastic material, and it shows the highest content of biogenic pelagic silica as approximated by high SiO2 and SiO2/Zr values (Table 1) and lack of their correlation with refractory siliciclastic-residing elements bundled in TIP (see supplementary material). Pelagic origin of silica is confirmed by radiolaria occurring in dissolved calcareous nodules en masse (S. Gouwy in Kabanov et al. 2016c). Hyaline sponge spicules that occur in overlying and underlying units are nearly absent in the Dodo Canyon Member, indicating closure of intermittent windows for oxygen-based benthic life (Kabanov & Gouwy 2017). Anoxic horizons III (AH-III) and IV (AH-IV) occur at the base of the Dodo Canyon and at 12–14 m below the top, respectively. Gamma-ray proxies (total gamma ray, U(gAPI), Un) appear to trace these horizons in off-bank sections over hundreds of kilometres, as exemplified by their presence in the Rumbly Creek section (Fig. 4 and supplementary material). The Prohibition Creek Member of Loon Creek O-06 and Little Bear N-09 wells shows EFMo levels increasing to, and even exceeding, the values at four main anoxic horizons, but EFU signatures are weak (Fig. 4 and Table 1). This U signature staying at the background of the Prohibition Creek–Dodo Canyon interval does not give a chance to trace this high-Mo horizon with spectral and conventional gamma-ray logs, neither does it find geophysical or lithological correspondence in the laterally equivalent grey-shale facies of the Bell Creek Member. This high-Mo level in the Prohibition Creek Member along with AH-II, -III and -IV are depleted in siliciclastic material (depressed TIP) and are enriched in pelagic silica. This high-silica and low-TIP signature is weak to non-existent at AH-I.
Anoxic horizons cannot be traced in condensed Canol sections draping carbonate banks of the Kee Scarp Member (Fig. 2b; Kabanov 2017). The type section of the Canol Formation at Powell Creek (Fig. 3) has a thickness of 28 m and exemplifies the type of BAT sections where carbonate banks are missing and the Canol Formation often directly onlaps the Carcajou marker. Also at Powell Creek, a thick (4.0 m) unit of laminated calcisiltites deposited at the toe of an adjacent carbonate bank (allochthonous member; MacKenzie 1970; Pugh 1983) masks gamma-ray response in the Canol base. The overlying part of the Canol Formation is divided into the lower fissile mudrock segment with intercalations of graded allochthonous limestones, the middle non-calcareous and very siliceous unit (with high U(gAPI) and Un, and depressed CGR(gAPI)), and the upper dolomitic mudrock portion with increasing CGR(gAPI) (Kabanov et al. 2016b). The lower calcareous part at Powell Creek is correlated with the Vermillion Creek Member of the SOB area. AH-III and AH-IV of the SOB are thought to merge at Powell Creek into the middle part of the Canol Formation. The upper part of the Canol Formation in the type section correlates with the upper Dodo Canyon Member above AH-IV and may also include the offshore equivalent of the Mirror Lake clinoform and some part of the overlying Loon Creek Member.
Age and correlation of anoxic horizons
The summary of conodont biostratigraphy (Kabanov & Gouwy 2017) assigns the basal part of the Bluefish Member to the uppermost Eifelian ensensis Zone, which is confirmed by the most recent findings (Gouwy 2017a, 2018) and the summary of data from Hume River (Uyeno et al. 2017). Samples assigned to the ensensis Zone at Rumbly Creek Tributary lie just below AH-I as detected by spectral gamma-ray logs (Kabanov et al. 2016a). Polygnathus hemiansatus (index taxon for the base of the Givetian; Walliser et al. 1995; Becker et al. 2012) has not yet been found in the area, and the conodont association from just above the top of the Bluefish Member at Dodo Canyon occurs within the timorensis Zone (Gouwy 2016a). This places AH-I at, or close to, the Eifelian–Givetian boundary (Fig. 1). The regional suppression of the Hume carbonate shelf with anoxic waters and the corresponding deposition of the Bluefish black shale is thus the time equivalent of the Kačák anoxic event recorded in European basins as well as the basal anoxic horizons of the Marcellus Shale of the Appalachian Basin (Fig. 1).
The base of the Ramparts platform member at Powell Creek is situated within the upper part of the ansatus Zone, and species indicating the hermanni and disparilis zones have also been identified in the lower part of the Ramparts Formation (Uyeno 1978). The index species, Skeletognathus norrisi, appears near the top of the Ramparts (Uyeno 1978) within the dark argillaceous limestone placed by MacKenzie (1970) into the base of the allochthonous limestone but recently reassigned to the Ramparts Formation, probably its Carcajou Member (Kabanov et al. 2016a). The sample from the top of the allochthonous limestone (sensu Kabanov et al., 2016b) also yielded S. norrisi, and two samples, 4 and 5 m above, contain conodonts of the soluta–rugosa Zone, giving a range of Lower Frasnian (Gouwy 2016b). This pins the Givetian–Frasnian boundary at, or just above, the allochthonous limestone in the type Canol section. Unfortunately, assemblages from two samples collected from the Prohibition Creek outcrop (SOB area), from the base of the Canol Formation and just below it, do not narrow the age better than the upper disparilis–transitans interval (i.e. the entire Upper Givetian and Lower Frasnian; Gouwy 2017b). AH-II, however, occurs above these samples, suggesting an earliest Early Frasnian age where it probably correlates with the Frasnes anoxic shale of Western Europe, the lower Duvernay of the Western Canada Sedimentary Basin and the geographically extensive ‘Frasnian hot shale 1’ of North African basins (Fig. 1 and references in figure caption).
A juvenile specimen of Palmatolepis punctata was identified at Powell Creek 1.0 m below the sharp SGR and U(gAPI) kick correlated with the base of Dodo Canyon Member (Gouwy 2017b). It is assumed that this first occurrence indicates the base of the Middle Frasnian, although fully formed elements of P. punctata are required for confirmation. The base of the punctata Zone is also indicated by the first occurrence of P. aequalis at the base of the Dodo Canyon Member at Prohibition Creek (Gouwy 2016b). At the Norman Wells quarry, the top of the Kee Scarp Member approximates the Lower–Middle Frasnian contact with the top of Kee Scarp containing a transitans Zone assemblage and the base of the overlying Canol Formation containing Pa. punctata (Kabanov & Gouwy 2017).
Age constraints for the upper part of the Canol Formation and its top are available from only Powell Creek (Gouwy 2017b). The upper portion of the Dodo Canyon Member, approximately matching AH-IV, lies within the lower hassi Zone. Three closely spaced samples from nodular carbonates at the top of the Canol Formation yielded a jamieae–lower rhenana conodont assemblage (Gouwy 2017b). Based on these age brackets, AH-III and -IV fit into the Middle Frasnian punctata and hassi zones, which are also known to contain major source rock units globally such as the Muskwa–Duvernay–Ireton of the Western Canada Sedimentary Basin, Middlesex–Rhinestreet of the Appalachian Basin, and the Domanik of the Timan–Pechora and Volga–Uralian hydrocarbon provinces of Russia (Fig. 1).
Geochemical signatures of palaeoceanographic regime
The following three groups of lithogeochemical proxies are discussed here: (1) the pattern of U–Mo covariation as a proxy for the oceanographic openness or restriction of an anoxic depositional system (Algeo & Tribovillard 2009); (2) patterns of P and TOC as a proxy for the primary planktonic production of organic matter (Algeo & Ingall 2007; Schoepfer et al. 2015); (3) Zn, Cu and Cd that co-precipitate with Mo and U in the anoxic sulphidic environment and are also micronutrients that are delivered to the sediment with dead organic matter (Borchers et al. 2005; Tribovillard et al. 2006; Böning et al. 2009).
Molybdenum and uranium
Analysis of the integrated elemental dataset of the study area (see supplementary material) reveals strong positive covariation of EFU and EFMo in black-shale units where marine signals are less influenced by terrigenous input compared with grey-shale units (Fig. 5). The Prohibition Creek Member (where it can be subset from the upper Hare Indian Formation), the Bluefish Member and the Canol Formation are uniformly enriched in Mo with EFMo/EFU ratios 3–4 times higher than the reference level for present-day seawater (SW; Algeo & Tribovillard 2009). The basal Imperial Formation is dominated by organic-rich laminated shales and siltstones, and its EFMo/EFU trend is lower, but otherwise similar to, that for black shales of the Horn River Group (Fig. 5d). The most offshore Trail River locality shows even higher EFMo/EFU values (Table 1 and Fig. 5c). The Road River–Canol Transition Zone unit (Fraser & Hutchison 2017) shows an offset resembling siliciclastic trends in grey shales and siltstones of the upper Hare Indian Formation (Fig. 5a and b).
Mo/U enrichment factor crossplots in mudrock units, based on 1687 ICP samples (see supplementary material). Trail River data are shown in yellow. Green field in (a) outlines the oxy-hydroxide particle shuttle systems (dominated by Cariaco Basin data), and grey area outlines the upwelling-affected Pacific continental margins with anoxia growing from suboxic (pale) to euxinic (dark), based on modern basins (Tribovillard et al. 2012).
The Road River–Canol Transition Zone unit of Fraser & Hutchison (2017) is a thin (2.3 m), very condensed, dark grey shale with elevated Ni concentration, hosting barite nodules, and is less organic rich than the Canol Formation. Fraser & Hutchison (2017) suggested that the Road River–Canol Transition Zone unit may be correlated with the uppermost Hume Formation, the entire Hare Indian Formation and the lower Ramparts Formation, which makes its geochemical imprints different from those of any unit in the study area. The Road River–Canol Transition Zone unit has facies, and probably age, counterparts in the nickeliferous, hyper-enriched in metals, black-shale horizons of the southern Richardson Basin and adjacent Selwyn Basin (Gadd & Peter 2018).
Phosphorus and TOC
TOC and P do not covary significantly in the studied sections. Pearson r is slightly higher in anoxic siliciclastic-lean units than in subsets dominated by siliciclastic material (correlation tables in supplementary material). The EFP notation inverts the pattern towards better correlation in siliciclastic-rich units, but r remains low except for the upper Hare Indian subset. In the basinal black-shale sections of the Horn River Group, the regression values of P fluctuate well below the average shale value of 0.07% (Fig. 6), with rare spikes above 0.1%, and attaining a maximum of 0.14% in the Canol subset (Table 1). The variance of EFP almost entirely accounts for the fluctuation of Al, making the Trail River section appear slightly more enriched in P than the two sections of the SOB area (Fig. 6). A spike of P, still below the reference PAAS value for shales, is traced in the base of the Canol Formation, immediately underneath AH-II (arrows in Fig. 6). This spike corresponds to a negative excursion in EFMo and EFU (Fig. 4). However, two other horizons of receding EFMo and EFU levels at the upper Bluefish–Francis Creek and the upper Vermillion Creek intervals (Fig. 4) do not coincide with traceable P signals.
P–TOC stratigraphy of Horn River Group in three sections shown as LOESS regression curves with α-tension 4%. The Trail River lithology is interpreted from Fraser & Hutchison (2017); subdivision is retained from the same source. RCTZ (here and in Figs. 7 & 8), Road River - Canol transitional zone (Fraser & Hutchison, 2017). (See Fig. 4 for colour-coded lithology, and supplementary material for data scatterplots.)
Zinc, cadmium and copper
Zinc and cadmium show enrichments in the Bluefish, Dodo Canyon and Prohibition Creek members relative to intervening siliciclastic-rich units (Table 1), but this enrichment with median EFs <10 for both elements is considered weak. EFMo/EFZn and EFMo/EFCd ratios are generally low (Figs 7 and 8). Enrichment of Cu is even lower than correspondent values of Zn (Table 1 and supplementary material). In the Canol Formation, the Dodo Canyon Member from sections located in the Mackenzie Valley, in relative proximity to the Laurentian sourceland, shows higher Zn enrichment and a better EFMo/EFZn correlation than the samples from Trail River (Fig. 8).
Mo/Cd enrichment factor crossplots in mudrock units cut off at 0.95 percentile on both axes. Data subsets are identical to those in Figure 5. Reference fields from Sweere et al. (2016): black regression line is sediments of the Gulf of California (Cd/Mo = 0.21) separating the high-productivity upwelling field from intermediate systems; red line is a cut-off at Cd/Mo = 0.1 for restricted basins. Grey bar denotes content in present-day marine phytoplankton, with Mo ≈ 2.0 ppm and Cd ∼ 3.2–22.0 ppm (Brumsack 1986).
Mo/Zn enrichment factor crossplots in mudrock units at 0.95 percentile on both axes; same subsets as in Figure 5.
Discussion
Could the Mackenzie Shelf be a restricted basin?
The search for possible Middle–Late Devonian oceanographic barriers as a factor in the development of basin stratification leads to the structurally complex region of the Paleozoic Selwyn Basin and the cordilleran terranes west of it (Fig. 2b). Based on the offsets of δ34S signatures of barite and pyrites collected from the Selwyn Basin, relative to the coeval average seawater values, Goodfellow & Jonasson (1984) interpreted the Selwyn Basin as a euxinic permanently stratified depositional system with impeded water recharge from the ocean. However, present-day knowledge on fractionation of sulphur isotopes in sulphide–barite deposits casts doubt on the euxinic model (Magnall et al. 2016a, b); in light of this, the openness v. restriction of the Selwyn Basin is in need of further discussion.
The Trail River section occupies the most offshore (westernmost) position in the Devonian palaeogeography and is most depleted in siliciclastic components and enriched in pelagic silica (Table 1). However, an increase in the proportion of sandstones and siltstones in the Canol Formation was recorded south of Trail River in the Ogilvie Mountains near the southwestern throat of the Richardson Anticlinorium (Fig. 2b; Norris 1997). Further SSW in the Selwyn Basin, the Portrait Lake and Misfortune formations of the Earn Group contain time equivalents of the Horn River Group (Turner et al. 2011; Gordey 2013). Both units are dominated by basinal mudrocks and chertstones but, in contrast to the Canol Formation, also contain lithic sandstones and conglomerates that form units of ≤200 m in thickness (Gordey & Anderson 1993; Turner et al. 2011). This clastic material was deposited in a deep-water submarine fan setting and was derived from older Selwyn Basin sediments with no evidence of remote sourcing (Gordey et al. 1987; Gordey & Anderson 1993; Gordey 2013). The Portrait Lake Formation is the host of Pb–Zn–barite deposits of the Macmillan Pass (Abbott & Turner 1991; Magnall et al. 2016a, b), one of a chain of nickeliferous, hyper-enriched in metals, black-shale occurrences (Gadd & Peter 2018). These massive sedimentary–diagenetic hydrothermal deposits record deep basin venting linked to the Late Devonian back-arc rifting (Gordey & Anderson 1993; Nelson et al. 2006, 2013). Conodont assemblages constrain Macmillan Pass deposits within hassi–Lower rhenana zones (Irwin & Orchard 1991), thus including the time equivalent of the upper Canol Formation. Western source areas of Earn Group clastic deposits represented a local chain of highs within a deep basin (Gordey & Anderson 1993). These highs, however, were not of nearly the same magnitude in clastic production as the coeval Ellesmerian Orogen of the Canadian Arctic (Lane 2007) or the Antler Orogen of the North American Cordillera in the USA and its extension in southern Canada (Root 2001; Colpron & Nelson 2009).
Most of the evidence for Middle–Late Devonian siliciclastic sourcing west of the Selwyn Basin comes from the Cassiar, Quesnellia and Yukon–Tanana terranes juxtaposed against the Selwyn Basin along the Tintina Fault, although none of these terranes had been identified as a source of sediments in the Selwyn Basin (Fig. 2b; Cecile & Norford 1993; Nelson et al. 2006; Colpron & Nelson 2009). This assemblage bears Late Paleozoic volcanic arc and back-arc features and is reconstructed in a para-autochthonous position to the ancestral North America prior to Late Devonian rifting (Nelson et al. 2006, 2013; Colpron et al. 2007). The Fairbanks Terrane (Fig. 2b) includes tectonic blocks of ancestral North America continental-margin affinity (Dusel-Bacon et al. 2006). Middle–Upper Devonian basinal and volcanogenic metasediments in these blocks include carbonate-platform units that met their demise at approximately the same time as the Hume carbonate platform of the Mackenzie Shelf. The Fairbanks Terrane is reconstructed as representing the former western side of the Selwyn Basin, before the onset of Late Devonian rifting (Dusel-Bacon et al. 2006).
The pericratonic nature of Cassiar, Quesnellia and Yukon–Tanana terranes is challenged by faunal, palaeomagnetic and palaeobotanical signatures of the Cassiar Terrane (Johnston 2008). In the interpretation of Johnston (2008), terranes now juxtaposed against the Selwyn Basin arrived from an exotic position close to the eastern Palaeotethys, accreted onto the ‘ribbon continent’ during the Triassic–Jurassic, and collided onto the ancestral North America in the Late Cretaceous, hence the possibility that the Selwyn Basin, or just the eastern part of it, was an open-ocean continental margin during the Devonian.
Mo–U covariation
Mo and U have long residence times in the modern ocean (0.78 and 0.45 myr, respectively) and therefore uniform distribution with a nearly constant Mo/U molar ratio c. 7.5–7.9 (3.023–3.184 weight ratio; Algeo & Tribovillard 2009). Dissolved Mo resides in oxic waters as largely unreactive oxyanion (MoO42–), but rebinds into thiomolybdates when (HS–) concentration exceeds a critical level (50–250 µM l–1 of HS–) during the development of euxinia. Transfer of Mo into sediments proceeds through adsorption of thiomolybdates on particulate organic matter and Mn- and Fe-oxyhydroxide particles (Algeo & Tribovillard 2009; Scholz et al. 2017). The latter form in the water column at the redox boundary, absorb Mo when sinking, and reductively dissolve in deeper anoxic waters releasing Mo close to the sediment surface, where it partly recycles into solution and partly co-precipitates with Fe sulphide minerals. This oxyhydroxide particle shuttle does not reach bottom waters in silled basins with permanent bottom euxinia but has the potential to greatly increase Mo concentration near the seafloor if the redox regime is mildly anoxic and/or frequently fluctuates as exemplified by OMZ-impinged continental slopes and weakly restricted basins with rapid water turnover (Algeo & Tribovillard 2009; Scholz et al. 2011, 2017).
Uranium behaves similarly to Mo in oxic and suboxic waters but markedly differently in anoxic waters. Reduction of U(VI) into reactive U(IV) commences at the redox boundary. U is not affected by the Mn- and Fe-oxyhydroxide shuttle and precipitates in the sulphate-reducing bacterial environment, mainly within the sediment, through interaction with particulate organic matter (Algeo & Tribovillard 2009). These differences, along with trace metal budgets, explain the variability of Mo/U ratios in modern oxygen-deficient basins and can be used to interpret redox conditions in the geological record. Open-ocean systems with mildly anoxic seafloor conditions will precipitate U ahead of Mo. Developing anoxia in an open system would precipitate Mo and U at equal intensity, with the Mo/U ratio of the sediment exceeding the Mo/U ratio of seawater. In a silled, stratified basin, the Mo–U covariation becomes imbalanced towards either Mo enrichment or its depletion (Algeo & Tribovillard 2009; Algeo & Rowe 2012).
The Horn River Group and basal Imperial EFMo/EFU values are higher than those in the abyssal-plain sediments of the Black Sea but similar to those from the Cariaco basin of the southern Caribbean, where a low sill allows for rapid water renewal with a complete turnover every 50–100 years (Tribovillard et al. 2012), and also to those of the Peruvian upwelling margin (Scholz et al. 2017). In the last example, sediment cores for the past 140 kyr show that Mo-enriched intervals (where EFMo/EFU ratios are higher than the SW value) match present-day seafloor areas characterized by nitrous anoxia, Mo fixation in a sulphidic environment just below the sediment–water interface, and strong Mn- and Fe-oxyhydroxide shuttling down the water column (Scholz et al. 2017). Based on data collected by Algeo & Tribovillard (2009), these EFMo/EFU values are higher than in most Devonian black shales of North America.
P and TOC as proxies for bioproductivity
The west-coast palaeogeographical location of the study area, just north of the equator, and the assumed openness to Panthalassa imply that the depositional system could have been affected by the upwelling if ocean circulation were operating in the actualistic mode. Phosphorus enrichment and phosphorite occurrence is a long-recognized fingerprint of high bioproductivity, particularly in association with upwelling systems (Föllmi 1996; Schoepfer et al. 2015). Phosphorus is effectively taken up by plankton, delivered into sediments with dead organic matter and rapidly remineralized by bacteria. Under strongly and permanently anoxic conditions, the released P diffuses back into the water column in a dissolved state (referred to as TPO4; Lomnitz et al. 2015) This creates the biochemical loop, but also strips sediments of P save for unreactive vertebrate skeletal and terrigenous particles. Recycling of P in anoxic waters may also be enhanced by the dissolution of P-carrying Fe-oxyhydroxides. The Corg/P ratio in strongly anoxic sediments usually exceeds the Redfield ratio of 106:1 (Algeo & Ingall 2007). Mechanisms of phosphorus retention emerge where O2 > 0 (i.e. in nitrous anoxic and suboxic waters), and can greatly increase at the intrasedimentary redox boundary. The latter serves as a trap for TPO4 emitted from underneath the redox boundary (Ingall et al. 2005; Algeo & Ingall 2007; Arning et al. 2009). Therefore, P uptake and the growth of phosphate nodules is detected under OMZs in suboxic bottom waters or where the redox barrier seasonally fluctuates across the sediment–water interface.
Siliciclastic-lean mudrock units of the Horn River Group are depleted in total P (Fig. 6). The only spike that may be described as a weak enrichment occurs in the base of the Vermillion Creek Member (arrows in Fig. 6) where it corresponds to a negative excursion in EFMo and EFU (Fig. 5). This suggests a milder anoxic to hypoxic regime, potentially suitable for some P retention in sediments. The overall phosphorus-lean character of the studied sections is consistent with qualitative palaeontological observations indicating the paucity of fish remains and conodonts, as well as the absence of any reports indicating phosphorite occurrence in time-equivalent strata anywhere across the Mackenzie Shelf. The condensed sections and hardgrounds at the drowning unconformity in the top of Hume Formation were never found affected by phosphatization either (Kabanov & Gouwy 2017).
This phosphorus-lean character extends westward to condensed shale sections of the Selwyn and Richardson Basins hosting nickeliferous, hyper-enriched in metals, black-shale horizons (Gadd & Peter 2018). Fernandes et al. (2017) reported the only confirmed presence of a phosphatic mineral film in the barite horizon. The latter is traced in the upper Canol Formation in the eastern Selwyn Basin and the adjacent MacMillan Pass mudrocks hosting massive sulphide-barite deposits. Low TOC/P ratio of 30–40 reported from Macmillan Pass might represent a signature of P retention in sediments (Magnall et al. 2018), but seems insufficient to make a strong argument for high rates of primary production. Overall, a hydrothermal-vent supply of Ba and associated metals seems more credible (Magnall et al. 2016a, b).
On a secular Phanerozoic trend proposed by Algeo & Ingall (2007), the TOC/P ratio of studied black-shale units occurs at a median of 140 ± 77. This is higher than the Redfield ratio and present-day marine anoxic sediments, but much lower than in eight other examples of Middle Devonian–lower Mississippian black shales dominated by planktonic type II kerogen (Algeo & Ingall 2007).
Zn, Cu and Cd as proxies for bioproductivity
The need to find other indicators of bioproductivity arises from the fact that the paucity of P in anoxic sediments is not necessarily supply-limited but may be explained by loss of recyclable P from permanently anoxic seafloors. Barium is another biogenic element delivered to the sediment with dead organic matter and is customarily considered a proxy for palaeoproductivity (Schoepfer et al. 2015). However, the use of Ba is limited by its intrasedimentary migration below the sulphate–methane boundary, which redistributes original dispersed barite into diagenetic concretions (Henkel et al. 2012). This is clearly seen in the Horn River Group, where Ba does not covary with other potential productivity proxies and instead forms random enrichment spikes apparently related to diagenetic concentrations (see supplementary material). In contrast to Ba, bioessential metals Zn, Cu and Cd precipitate as sulphides, which immobilizes them in a sulphidic diagenetic system (Huerta-Diaz & Morse 1992; Calvert & Pedersen 1993; Tribovillard et al. 2006).
Zinc and copper are micronutrient metals taken up by plankton in surface waters and shuttled down the water column with dead organic matter and with (Fe, Mn) oxyhydroxides, leading to greatly enhanced bottom water concentrations in modern marine systems. Under euxinic conditions, Zn and Cu may precipitate as an isomorphic phase in pyrite or, to a lesser degree, their own mineral phases (Calvert & Pedersen 1993; Tribovillard et al. 2006). However, modern high-productivity systems of OMZ-impinged oceanic shelves show only weak to moderate enrichment in authigenic Zn and Cu, indicating that these metals may recycle into seawater through pathways that are currently poorly understood (Borchers et al. 2005; Böning et al. 2009).
Cadmium distribution in modern oceanic waters tends to mimic that of micronutrient metals, particularly Zn, indicating its bioessentiality, although increased concentrations of Cd are toxic for phytoplankton as well as for higher consumers. Cadmium is thought to be involved in metabolic reactions in phytoplankton under limitation of Zn and Mn (Xu & Morel 2013). Enrichment in authigenic Cd is very pronounced in sediments of modern highly productive open-shelf systems (Fig. 7), which justifies its use as an indicator of palaeo-bioproductivity (Sweere et al. 2016). However, this usefulness of Cd is limited by the evolutionary trend in marine phytoplankton (Riegel 2008). Modern phytoplankton are volumetrically dominated by the ‘red plastid lineage’ represented by dinoflagellates, coccolithophorids and diatoms. These groups evolved in the Late Triassic–Early Jurassic and thrive with ample phosphorus supply and the availability of Mn, Cd and Co. Early–Middle Paleozoic plankton were dominated by the ‘green lineage’ represented by prasinophytes and acritarchs. These groups preferred high nitrogen supply and utilized Fe, Zn and Cu as micronutrients with no indication of Cd uptake. The ‘green plankton’ and ‘red plankton’ epochs are divided by the ‘Late Paleozoic phytoplankton blackout’, which lasted from the decline of acritarchs across the Devonian–Carboniferous boundary until the Late Triassic (Riegel 2008).
In the Horn River Group, none of Zn, Cu or Cd build to concentrations that would be expressed in EFs persistently above 10 (Table 1) and none covary with Mo to a significant extent (Figs 7 and 8). Out of these three metals, Cu is the scarcest (Table 1; supplementary material) and is not discussed further. Variation of both EFZn and EFCd in the Canol Formation can be accounted for by siliciclastic influx from the Laurentian sourceland rather than oceanic supply of these metals. The Trail River section has very low concentrations of both elements (Table 1), and the SOB subset is relatively enriched in Zn (Figs 7 and 8).
Whereas weakness of Cd enrichment can be explained by inability of the Paleozoic phytoplankton to utilize this metal (Riegel 2008), lack of strong and persistent enrichment of Cu and Zn, depleted values of P and scarcity of skeletal phosphate indicate that the sedimentary system might have been overall oligotrophic and high TOC content reflects high preservation potential of deposited organic matter in combination with slow sedimentation rate, rather than high rates of primary production (Algeo & Ingall 2007).
Global controlling factors
Based on conodont biostratigraphy, anoxic horizons recognized in the Eifelian–Frasnian of the study area (Fig. 4) appear to correlate with time equivalents globally (Fig. 1). This suggests a global control on the occurrence of shelfal anoxia. The west-coast location of an open-shelf sedimentary system, absence of evidence for enhanced oceanic nutrient flux, and global expansion of anoxic sediments in shelfal basins can be reconciled through the elegant model of Kidder & Worsley (2004, 2010). This model recognizes three principal states of the World Ocean and sets greenhouse conditions as the default state accounting for no less than 70% of the Phanerozoic time. The icehouse mode, with its characteristic strong thermal circulation, lasted about 20–25% of the Phanerozoic. Thermohaline circulation in the modern icehouse ocean consists of thermal and haline components. The thermal mode is driven by strong temperature gradients between low and high latitudes characteristic of icehouse climates, caused by the sinking of cold brines in polar regions. The haline mode describes the sinking of warm, dense brines concentrated by evaporation, which mostly occurs in arid mid-latitudes. In addition to thermal and haline components, it is possible that geothermal heating from the ocean floor is capable of solely driving circulation, although at much slower rates of c. 15 000 years for complete turnover (Worthington 1968) (compared with today's 1000–2000 years under thermohaline regime; Böös et al. 2012).
Global cooling into icehouse mode is correlated with enhanced silicate weathering in humid equatorial regions through the fixation of ocean-atmospheric CO2 into carbonate sediments with ample newly released Ca2+ and Mg2+ (Walker et al. 1981; Kent & Muttoni 2013). The greenhouse ocean is characterized by weakened thermal overturn in polar regions and, consequently, less oxygen injection into the deep ocean, leading to the expansion of OMZs. Expanded anoxic conditions, in turn, promote the loss of nitrogen from the biochemical cycle into the atmosphere in the form of N2 and N2O. The hothouse is a disastrous, usually short-living (≤1 myr, rarely up to 3 myr) planetary condition that accounts for <4% of the Phanerozoic time and is frequently coincident with mass extinctions (Kidder & Worsley 2010). The hothouse Earth features ice-free polar regions, weak oceanic haloclines and haline-mode ocean convection that drives polar upwelling of warm anoxic waters. As no sinking of cold brines occurs, the water turnover is sluggish, allowing time for geothermal heat to warm bottom waters. The hothouse is an unstable condition requiring forcing mechanisms, which are seen in plume-head degassing associated with continental large igneous provinces (LIPs). Such degassings, predating or accompanying main basalt outflows, can rapidly release enough CO2 to cause global warming and outweigh cooling effects from co-erupting SO2 and mineral aerosols (Bond & Wignall 2014). The OMZ expands dramatically under hothouse climates and transfers into an euxinic state while its upper boundary shallows to the photic zone.
Following the initial idea of Kidder & Worsley (2010), it is postulated that the origin of the Kačak Event lies in the dramatic expansion of OMZs, resulting in flooding of lower shelves and epeiric basin interiors with anoxic waters. Modern shelves do not provide analogues for this condition as no epeiric basins are facing upwelling-driven OMZs, but this scenario has been employed to explain spreads of Late Pennsylvanian phosphatic black shales in the midcontinent USA seaway during icehouse interglacials (Algeo & Heckel 2008). Givetian–Frasnian greenhouse conditions must have been more conducive for shallower OMZs than today's 100–900 m (Gilly et al. 2013), hence the widespread occurrence of anoxic bottom waters in shelfal basins. Mechanisms of persistent stratification in these basins would have involved a quasi-estuarine factor by which riverine influx of freshwater kept the surface layer lighter, as well as greatly reduced thermal oceanic convection with attenuated recharge of nutrients as a consequence. This latter explains the oligotrophy of the Mackenzie Shelf. Yet another factor potentially accounting for persistent stratification is seen in the predicted wane of wind speed with progressive global warming (0.83 of the present-day value in the greenhouse and 0.67 in the hothouse; Kidder & Worsley 2010). This calming would contribute to a slowdown of wind-driven oceanic currents, greatly diminish physical weathering and reduce aeolian delivery of iron and other nutrients to surface waters, thus contributing to oligotrophy. At the same time, Kidder & Worsley (2010) speculated that cyclonic storms will probably strengthen and reach out to high latitudes with progression of global warming towards hothouse conditions. These cyclonic storms will push the storm wave base to greater depths, entrap euxinic waters and release H2S into the upper productive oceanic layer and the atmosphere.
In the overall ‘warm greenhouse’ (sensu Kidder & Worsley 2012) stratified basins of the Givetian–Frasnian time, globally correlated anoxic horizons may correspond to hothouse crises in their nascent or ‘aborted’ form (Kidder & Worsley 2010). This underdevelopment follows from a lack of major biotic perturbations, except for the Frasnian–Famennian mass extinction and less severe Taghanic biocrisis, as well as positive δ13C isotope excursions associated with Devonian black-shale events (Fig. 1). A series of minor biotic perturbations associated with pre-Taghanic black-shale events were not nearly the same magnitude as the Frasnian–Famennian event (Bond & Grasby 2017). A fully developed hothouse crisis is predicted to show a negative δ13C swing caused by a drop in primary production as a result of nutrient arrest in expanded OMZs (Kidder & Worsley 2010; Grasby et al. 2016). Some horizons of fully developed OAEs also record negative excursions of Mo, U and Zn, which are probably a response to drawdown of oceanic inventories of these metals by enhanced precipitation in expanded anoxic waters (Algeo & Rowe 2012). These negative swings of trace metals are not observed in the Horn River Group. A pilot indication of the hothouse nature of the Frasnian–Famennian event is available from the negative δ15N excursion reported by Levman & von Bitter (2002). Until further clarification, it is assumed that positive δ13C excursions were driven by pulses of degassing in Devonian LIPs masking possible drops in bioproductivity. This vision of several ‘nascent hothouse’ episodes in the latest Eifelian–Frasnian, however, is not consistent with the low-latitude sea-surface temperature model from δ18O of conodont apatite (Joachimski et al. 2009). This model depicts the Early Givetian as being a time of cooling to the present-day low-latitude level of around +20°C followed by steady increase to its culmination at +30–32°C in the latest Frasnian (Fig. 1).
Pulses of anoxia imprinted in four recognized anoxic horizons are coupled with the attenuation of siliciclastic components (Fig. 4). If these pulses were indeed hothouse developments, sensu Kidder & Worsley (2010), sourceland retreats may be tied to thermal eustatic transgressions, in compliance with predicted thermal sea-level rise by 10–13 m per 10–12°C of ocean warming (Miller 2009). Thermal expansion of the oceanic water mass, however, does not offer a mechanism for amplitudes of sea-level change exceeding 15–20 m, whereas other causes of eustasy in the ice-free Middle–Late Devonian Ocean (e.g. mid-ridge pulsation; Johnson et al. 1985) remain purely speculative. Therefore, scrutiny of the Devonian eustatic curve (Johnson et al. 1985) is required to check if, and to what extent, sea-level changes were ‘major’, as postulated (Sandberg et al. 2002; Arthur & Sageman 2005), and whether the fluctuation of the redox boundary in the stratified ocean offers a better explanatory platform for Devonian cycles than genuine sea-level changes.
LIPs as a potential trigger of Devonian anoxic events
Continental LIPs, potentially accountable for the spread of black shales and the Frasnian–Famennian biotic crisis, occur in Eastern Europe and Siberia (Fig. 1; Kravchinsky 2012; Bond & Wignall 2014). Trap basalts of the Viluy Rift in Siberia were emplaced 380–340 myr ago based on absolute dating, with the major phase of lava flow occurring at 370–360 Ma (Courtillot et al. 2010). The total volume of Viluy basalts remains unknown, with a rough estimate of c. 1 × 106 km3 (Courtillot et al. 2010). The Pripyat’–Dnieper–Donets rift system is estimated to have 1.5 × 106 km3 of basalts that erupted over a relatively short period of time bracketed within 367–364 Ma (Kravchinsky 2012). The alkaline magmatic complexes of the Kola Peninsula are dated within the range 385–363 Ma (Kravchinsky 2012; Wu et al. 2013). Effusive rocks in the Kola LIP are not preserved but are unlikely to be as voluminous as in the other two LIPs (Bond & Wignall 2014). Available age ranges in LIPs appear to be younger than the Kačak spread of anoxic shales, and most ages are younger than the basal Frasnian anoxic horizon (Fig. 1). Assuming that age interpolations of stage boundaries are ultimately correct in the Geological Time Scale 2012 (Becker et al. 2012), this offset in geological age can be explained either by mantle-plume degassing starting earlier than eruption activity or an incompleteness of dated materials from LIPs. The latter explanation is favored as new data tend to push LIP ages back in geological time (Wu et al. 2013; Bond & Wignall 2014).
Conclusions
The age of the Horn River Group is constrained with updated conodont identifications between the latest Eifelian and the early Late Frasnian. During this time interval of c. 13 myr (Becker et al. 2012), a pronounced facies zonation developed through the deposition of Laurentian-sourced deltaic lobes prograding into the stratified basin with anoxic bottom waters. These lobes formed topographic highs above the chemocline and became outgrown by isolated carbonate banks as siliciclastic influx waned across the Givetian–Frasnian boundary. The resultant lateral zonation in the preserved structural layout is described as the southern and western off-bank areas (SOB and WOB; strongly dominated by basinal organic-rich mudrocks) and the bank-and-trough area (BAT; thick grey shales, siltstones, shallow-water carbonates and thin basinal mudrocks).
Aluminium-normalized Mo and U logs (706 data points, enrichment factor notation) reveal horizons of enhanced anoxia (anoxic horizons; AHs) in two superficially monotonous black-shale cores of the SOB area. These horizons show enhanced content of pelagic silica and attenuated siliciclastic components. Traceability of AHs in other surface and subsurface sections of SOB and WOB areas is tested with spectral gamma-ray components K + Th (siliciclastic input), and Un (trace-metal enrichment). AH-I, located in the middle of the Bluefish Member, approximates the Eifelian–Givetian boundary and correlates with the Kačak global black-shale event. AH-II, in the lower part of the Vermillion Creek Member (lower Canol Formation), coincides with extensive black shales of earliest Frasnian age, usually referred to as the Frasnes event. The Dodo Canyon Member of the upper Canol Formation hosts AH-III and AH-IV. AH-III marks the base of the Dodo Canyon Member and is held accountable for the ultimate demise of Ramparts carbonate banks. The possibility exists, however, that the tallest carbonate banks, found in the subsurface of the Peel Plateau, survived through the Canol and became buried under the fine-grained terrigenous sediments of the Imperial Formation. AH-III and AH-IV fit into the entire Middle Frasnian and have time equivalents in major global anoxic shales (AH-III = Middlesex, AH-IV = Rhinestreet in the Appalachian Basin; AH-III = punctata δ13C excursion in Western Europe; undivided Domanik, or its lower two-thirds, in Eastern Europe and the Urals). An increase in EFMo found between AH-I and AH-II may correspond to the Geneseo black-shale event of the Appalachian Basin and its correlatives. This AH-II and the high-Mo level between AH-I and AH-II do not have obvious correlatives in BAT sections, but the question remains as to whether AH-II is traceable within the Ramparts Formation as the Carcajou gamma-ray marker.
Clues to the cause of basin stratification are provided by palaeogeographical constraints of the studied strata and by Mo/U covariation in the elemental dataset of 1687 samples. Lack of an effective oceanographic barrier is supported by the overall siliciclastic-lean character of the Horn River Group and by the attenuation of siliciclastic components in the study area westward of 130°W. Further west and SW in Selwyn Basin, time-equivalent strata are predominantly basinal cherty mudrocks, but they contain lithic sandstones and conglomerates derived from local sources. The argument was also made that Selwyn Basin could represent the continental margin (Johnston 2008) rather than a pericratonic basin (Colpron et al. 2006, 2007), thus unequivocally placing the study area in an open-shelf system. Strong Mo/U covariation in basinal mudrock units and a westward Mo enrichment trend are in agreement with an unrestricted ocean-shelf water exchange system.
Signatures of oceanographic openness co-occur with evidence of nutrient limitation. Depleted total P, a lack of phosphorites in the surveyed area (including deep-water time equivalents in Selwyn Basin), paucity of fish and conodont remains, weak enrichments of Zn and Cd and their overall westward depletion from the SOB to the Trail River section indicate that upwellings were not operating as a factor of nutrient recharge. Oceanographic openness, nutrient limitation and global spreads of black shales in shelfal basins (Fig. 1) are reconciled through a model of Kidder & Worsley (2010). A shift to black-shale deposition occurred in the mid-Devonian in many shelfal basins of the World (Kačak Event), which can be explained by the expansion and shallowing of OMZs rather than genuine sea-level rise. Slowdown in ocean turnover, attenuation of west-coast upwellings and OMZ expansion are the signatures of the ‘warm greenhouse’ Earth mode. A trigger for this shift is seen in the onset of mass degassing in continental LIPs, most probably in the Vilyu Rift and Kola magmatic provinces. Discrete ‘black-shale events’ recognized in the latest Eifelian–Middle Frasnian are seen in this model as genuine Paleozoic OAEs marking hothouse crises sensu Kidder & Worsley (2010), probably in their nascent form as they do not bear the signatures of nutrient crises and drawdown of ocean inventory of Mo and U, and did not culminate in major mass extinctions like the one at the Frasnian–Famennian boundary.
Acknowledgements
This work is a contribution to the Mackenzie Project of the Geomapping for Energy and Minerals (GEM2) Programme, Geological Survey of Canada (Land and Minerals Sector, Natural Resources Canada) publication No. 20180027. Project administrative support has been provided by C. Ozyer, P. Wozniak and M. Francis. Peer reviews by A. Mort (GSC Calgary, Canada), D. P. G. Bond (University of Hull, UK) and an anonymous reviewer, along with editorial comments of G. Anthony Shields-Zhou (University College London, UK), resulted in profound improvements. S. A. Gouwy, R. B. MacNaughton and K. M. Fallas are GSC colleagues who provided expert advisory and field support. Those who offered their help with well log data, needed at the post-review stage, are M. L. Borrero (GSC) and C. Molaro (Husky Energy). The final stylistic check-up of the paper has been performed by R. B. MacNaughton and H. M. King (GSC). This work is also part of the IGCP-652 Project ‘Reading geologic time in Palaeozoic sedimentary rocks’.
Funding
This work was funded by the Natural Resources Canada (Mackenzie Project of Geomapping for Energy and Minerals Programme).
Scientific editing by Graham Shields-Zhou
- © 2018 Her Majesty the Queen in Right of Canada, as represented by the Minister of Natural Resources Canada. Published by The Geological Society of London. All rights reserved
References
Devonian (c. 388–375 Ma) Horn River Group of Mackenzie Platform (NW Canada) is an open-shelf succession recording oceanic anoxic events
