Outcrops with conspicuous reddish to yellow-colored clinker, blackish paralava, and blends of both with a breccia-like appearance occur across the Canadian Arctic. We examined such rocks on Ellesmere Island, Banks Island, and the Mackenzie Delta area. These rocks are a product from natural combustion of bituminous shale and low-rank coal seams in Cretaceous and Paleogene host sedimentary rocks, respectively. The main mineral phases of clinker and silicate paralava samples are comprised of quartz + hematite ± feldspars ± cristobalite (or tridymite) ± cordierite–sekaninaite ± clinopyroxene ± sillimanite ± glass. Slag-like iron oxide paralava (74–95 wt.% total Fe2O3) consisting of hematite ± magnetite ± clinopyroxene occur in Paleogene host sedimentary rocks, rich in siderite concretions. The whole-rock geochemical composition of clinker and silicate paralava shows similarities for samples from the same outcrop. Regional and local specific elemental enrichments are mainly inherited from the sedimentary protoliths, which are characterized by volcanogenic input (Paleocene sedimentary rocks) or oxygen-depleted depositional conditions (Upper Cretaceous bituminous sedimentary rocks). Spontaneous combustion could take place when the organic-rich sedimentary rocks become exposed to atmospheric oxygen. This process has occurred at least since the Messinian stage (Miocene) on Ellesmere Island (6.1 ± 0.2 Ma; 40Ar/39Ar incremental heating dating on whole-rock paralava) and continues until now. An active combustion process on scree from a coal seam and clastic Eureka Sound Group sedimentary rocks was observed on Ellesmere Island.

The spontaneous combustion of coal and organic-rich mudstones in natural outcrops and also related to human mining activities, is a common and long-known phenomenon around the world. It is triggered by the contact of such rocks to the atmosphere induced by geological processes. This causes the onset of exothermic oxidative reactions due to the presence of reactive macerals, such as huminite (Kalkreuth et al. 1996; Heffern and Coates 2004) or pyrite oxidizing to sulfate (e.g., Deng et al. 2015). Other factors such as particle size, moisture content, ambient temperature, etc. also influence the initiation of combustion (e.g., Grapes 2010). Spontaneous combustion is accompanied by combustion metamorphism (a special variety of pyrometamorphism) and the formation of combustion products. These include clinkers (thermally altered, brick-like, and partially vitrified pelitic–psammitic sedimentary rocks) and paralavas (fused sedimentary rocks with a dark, aphanitic, vesicular, and sometimes pahoehoe lava-like appearance), which are characterized by typical high-temperature/low-pressure mineral assemblages and glass (Cosca et al. 1989; Grapes 2006, 2010; Grapes et al. 2011).

In the Canadian Arctic, indications for the spontaneous combustion of coal (often lignite/sub-bituminous coal) or organic-rich dark mudstones were, amongst others, previously identified on Ellesmere Island (Figs. 1a and 1b). One location on Fosheim Peninsula was called “clinker” (Dawson et al. 1975). Two zones in a section north of Strathcona Fiord (Fig. 2) were labeled with the term “bocanne” (Ricketts 1994), which means “naturally burning shale banks” according to the definition of Crickmay (1967) suggesting burning shale in the outcrop. More recently, the occurrence of paralava and clinker in the Stenkul Fiord and Split Lake areas on southern Ellesmere Island (Fig. 1a) was reported (Piepjohn et al. 2007; Estrada et al. 2009). One coal-seam fire south of Stenkul Fiord has been dated to have burnt 3.3 ± 0.5 Ma ago, in the mid-Pliocene (40Ar/39Ar incremental heating dating on whole-rock silicate paralava; Estrada et al. 2009). Coal-bearing sedimentary rocks of the Paleogene Eureka Sound Group crop out in all of these areas.

The Smoking Hills stretching along the east coast of the Bathurst Peninsula into the Beaufort Sea (Fig. 1a) on the western Canadian mainland, was named by John Richardson during John Franklin’s second polar expedition in 1826 (Franklin 1828) and are also long known as Ingniryuat by Inuvialuit. Here, dark Upper Cretaceous shales have been continuously burning since at least then (Mathews and Bustin 1984; Grasby et al. 2022). Auto-combustion could have conceivably occurred since the Smoking Hills Formation was first exposed during regional uplift and erosion of overlying sedimentary rocks in the Miocene (Lane and Dietrich 1995). Clinker deposits have been reported in age equivalent Kanguk Formation rocks on Banks Island (Smith et al. 2016), as well as on Eglinton Island (Plauchut and Jutard 1976). A description of a scorching outcrop of nearly age equivalent oil-prone Upper Cretaceous rocks of the Boundary Creek Formation on the Yukon North Slope is given in Fraser and Reinhardt (2015).

A long-known example, slightly further south, is smoldering outcrops of coal on the western bank of Mackenzie River, south of Fort Norman (Fig. 1a), which was reported by Alexander Mackenzie who visited the place in August 1789 (Mackenzie 1801; Hume 1954). Here, sedimentary rocks of the Upper Cretaceous–Paleocene Summit Creek Formation with interbedded low-rank coals have been burning for centuries (Dixon 1999; Fallas et al. 2013).

In the same area, some kilometers south of Fort Norman, but about 30 km to the west of the Mackenzie River near Tate Lake, the “Tertiary Hills Clinker” were described (Fallas et al. 2013; Kristensen et al. 2019). The “Tertiary Hills Clinker” are outcrops of combustion-metamorphosed sedimentary rocks of the Summit Creek Formation, which were exploited by hunter–gatherers over the past 10 000 years. Porcellanite, a distinct form of clinker with a ceramic or porcelain texture, was collected as raw material to produce tools. These artefacts were found over a large area in archaeological sites, indicating a circulation from the outcrop in the Northwest Territories to the Yukon, northeast British Columbia and Alberta (Le Blanc 1991).

Another more distant example is known from western Greenland, where the self-combusting shales of the Albian–Cenomanian to Campanian Atane Formation crop out. These shales occur at several localities, but are especially common on Nuussuaq Peninsula (Figs. 1a and 1b) in the Paatuut and Kingittoq areas (Dam et al. 2009).

New manifestations of active and extinct natural combustion processes were coincidentally found on southern Ellesmere Island, Banks Island, and in the Mackenzie Delta area (Figs. 1a and 1b; Table 1) during geological field work related to the Circum-Arctic structural events (CASE) expeditions of the German Federal Institute for Geosciences and Natural Resources (BGR) with the Geological Survey of Canada (GSC) and the Yukon Geological Survey (YGS) between 2011 and 2016. Here, we present brief descriptions of 10 sites of combustion metamorphism, including a site with an actively burning fire on southern Ellesmere Island observed in summer 2011 (Fig. 1a). The resulting combustion-metamorphosed rocks appear very similar in the field, despite a distance of about 2000 km between the northeasternmost and the southwesternmost outcrops (Fig. 1a), different stratigraphic positions of the sedimentary host rocks (Fig. 1b), and different fuel sources for spontaneous combustion (lignite/sub-bituminous coal seams in Paleogene sedimentary rocks from Ellesmere Island and Mackenzie Delta and dark, organic–rich mudstones in Cretaceous sedimentary rocks from Banks Island). From each outcrop, one typical clinker sample and, wherever present, one or two typical paralava samples were analyzed for their whole-rock mineralogical and geochemical composition. Scanning electron microscopy was used to characterize the internal texture of selected paralava samples. 40Ar/39Ar incremental heating dating on whole-rock paralava samples provided insight in the respective age of combustion.

Our scientific aims are to provide indications supporting the identification of combustion metamorphic rocks in the field; to provide mineralogical and geochemical characterization of paralava and clinker to compare the compositions in the same outcrop and on a regional scale; to discuss relationships between the combustion–metamorphic rocks and their original depositional environment; and to examine the relationship between the age of natural combustion events and geomorphological processes. This allows insights into the potential of these surface-related rocks derived from distinct lithologies to inform about the onset of landscape-forming processes in periods lacking often any other sedimentary record.

The identification of rare combustion–metamorphic minerals, mineral–chemical analyses, and modeling of pressure and temperature conditions during combustion metamorphosis and melting processes are beyond the scope of this overview study.

Despite the wide distances between outcrops described herein from Ellesmere Island, Banks Island, and the Mackenzie Delta area (Figs. 1a and 1b), these regions share a common geological history reflecting the development of the North American continental margin since Precambrian times.

Crystalline basement rocks of the Archean to Mesoproterozoic Greenland-Canadian Shield form the oldest rock units (e.g., Frisch and Trettin 1991; Trettin 1991a). Passive margin sedimentation took place in the Franklinian Basin between Neoproterozoic to Devonian times (e.g., Thorsteinsson and Tozer 1957; Stuart Smith and Wennekers 1977; Dewing and Hadlari 2023). In the late Devonian until the earliest Carboniferous, Ellesmerian deformation affected the northern part of the Franklinian Basin from Ellesmere Island in the NE to Mackenzie Delta in the SW (Thorsteinsson and Tozer 1957, 1970; Norris and Dyke 1987; Coflin et al. 1990; Trettin 1991a, 1991b; Harrison 1995; Lane 2007; Piepjohn et al. 2015).

Following this deformation event, sedimentation continued in the Sverdrup Basin, a 1300 km by 350 km basin that extended from Ellesmere Island in the NE to Banks Island in the SW, and accumulated up to 13 km of sedimentary rocks of Carboniferous to Paleogene age (e.g., Embry and Beauchamp 2008, 2019). During the Cretaceous, mainly in the time span from ca. 130 to 75 Ma, the Sverdrup Basin was repeatedly affected by intense extrusive and intrusive magmatic activity of the High Arctic Large Igneous Province (HALIP). This was related to the beginning of the opening of the Amerasia Basin of the Arctic Ocean (e.g., Embry and Osadetz 1988; Buchan and Ernst 2006; Estrada et al. 2016; Bédard et al. 2021a, 2021b). Renewed volcanic activity took place with the opening of the N Atlantic (and the Eurasia Basin) and Baffin Bay during the Paleogene (e.g., Estrada et al. 2010; Larsen et al. 2016; Larsen and Williamson 2020).

Deposition in the Sverdrup Basin ended during the Paleogene as a consequence of regional compression, strike–slip movements and widespread uplift during the Eurekan Orogeny (Embry and Beauchamp 2019). Overall, strata in the Sverdrup Basin are deformed due to several factors, episodic flow of Carboniferous evaporites during the Mesozoic (Balkwill 1978; Boutelier et al. 2011; Galloway et al. 2013; Dewing et al. 2016), Barremian to Cenomanian magmatism and faulting (Embry and Osadetz 1988; Embry 1991), and compression during the Eurekan Orogeny in the Eocene. The Eurekan Orogeny produced high amplitude folds and thrust faults on Ellesmere and Axel Heiberg islands and smaller folds, e.g., on Melville Island (e.g., Miall 1986; Harrison et al. 1999). During this time, small basins with Paleocene to early Eocene deposits of the Eureka Sound Group and its equivalents were formed (e.g., Mayr and de Vries 1982; Miall 1986; Harrison et al. 1999). These sedimentary rocks comprise mainly material eroded from Paleozoic and Mesozoic rocks and minor volcaniclastic material related to coeval proximal and distal extrusive magmatism.

2.1. Ellesmere Island

2.1.1. Strathcona Fiord–Vendom Fiord area

The outcrops of combustion–metamorphic rocks examined here occur in Paleogene strata located in an area between Vendom and Strathcona fiords (Figs. 1a and 2). In this area, the Paleogene Eureka Sound Group comprises the fluvial to marginal marine Mount Bell, Mount Lawson, and the deltaic to nearshore marine Mount Moore Formations. The stratigraphically youngest unit of the Eureka Sound Group is the deltaic Margaret Formation (latest Paleocene to early Eocene) cropping out here in the center of a large syncline. Eureka Sound Group strata at Strathcona Fiord contain at least 16 coal seams with a maximum thickness of 11.4 m and a cumulative coal seam thickness of ca. 31 m (Kalkreuth et al. 1998). Moreover, various centimeter thick volcanic ash layers were identified in both Mount Lawson and Margaret Formation sedimentary rocks (Reinhardt et al. 2013).

Both the Paleozoic and Mesozoic strata as well as the Eureka Sound Group were strongly affected by strike–slip faults and thrusts related to the Eurekan deformation during the Eocene (e.g., von Gosen et al. 2012; Piepjohn et al. 2016). The youngest deposits of the area, the sedimentary rocks of the lower Pliocene Beaufort Formation (e.g., Tozer 1956; Fletcher et al. 2019), were not affected by Eurekan deformation, which ended at about 34 Ma (Piepjohn et al. 2016).

Locally, the sedimentary rocks of the Eureka Sound Group were affected by combustion metamorphism (see sample sites in Fig. 2). At one of these outcrops (site of samples CASE12_20 and CASE12_21), an active coal scree fire was observed in August 2011. The actual situation of this outcrop is described more detailed in the following.

2.1.2. Outcrop with actively burning coal scree

The burning coal scree fire was observed on the 24th August 2011 in the steep eastern canyon wall of a small river (Figs. 2 and 3). Here, sedimentary rocks of the Mount Bell and the Mount Lawson Formations are eroded. The sedimentary rocks are steeply inclined along a normal fault zone related to the Eurekan deformation. The exposed sedimentary rocks consist of weakly consolidated shale and fine- to medium-grained sandstone with several intercalated coal seams of decimeter-to-meter thickness. During the time the field team was present, fresh pieces of rock crumbled continuously from the steep canyon wall, caused by the thawing permafrost (air temperature during the day at nearby campsite of the CASE 12 expedition reached up to 16°C in August 2011) and fed the scree below.

Smoke rose from the scree at several positions all along an area of approximately 200 m × 50 m. Hot spots/pockets of embers, i.e., the glowing pieces of coal, were scattered all over the scree, some with continuous development of smoke, others without. In some places, embers were visible directly at the surface (Fig. 3). Grayish to whitish ash of the burnt coal pieces, clinker, and the combustion-metamorphosed pieces of shale and organic-rich sand, in color shades of brown, beige, and typical brick-red, surrounded the glowing coal pieces. Overall, the surface was hot, but could still be walked upon. Digging with a hammer revealed clinker pieces approximately 10 cm below the scree surface. Clinker was also present in areas that did not currently burn at the surface. The maximum depth of the burning fire could not be determined. Hot spots were only identified on the scree; neither of the coal seams were on fire. Neighboring areas of the scree, which lacked fresh material crumbling down the canyon wall were not burning either.

2.1.3. Bache Peninsula

The northernmost outcrop of paralava and clinker sampled in this study is present on Bache Peninsula, eastern Ellesmere Island near Bartlett Bay (Figs. 1 and 4). Here, the commonly unconsolidated Eureka Sound Group sedimentary rocks are preserved in a tectonic graben (Figs. 4 and 5), where downfaulting and subsequent partial inversion occurred during tectonic movements of the Eurekan deformation. Normal faults limit the west–east-oriented graben against lower-to-middle Ordovician limestones of the Franklinian Basin (de Freitas et al. 2007; von Gosen et al. 2019). Eureka Sound Group strata in the graben structure contain up to 38 coal seams with a maximum thickness of 4.7 m and a cumulative thickness ca. 56 m (Kalkreuth et al. 1993). A small amount of paralava and numerous clinker pieces were sampled.

2.2. Banks Island

Two outcrops with paralava and clinker were sampled in the southwest and northeast of Banks Island, respectively.

2.2.1. Nelson Head (South Banks Island)

In the outcrop KPA758 near Nelson Head (Fig. 6a) Lower Cretaceous (i.e., Isachsen, Christopher, and Hassel Formations) and Upper Cretaceous (i.e., Kanguk Formation) sedimentary rocks overlie Proterozoic basement, which crops out nearby along the coast. Further inland, sedimentary rocks of the Eureka Sound Group follow, which are unconformably overlain by Neogene sedimentary rocks of the Ballast Brook and Beaufort Formations (Fig. 1b; Miall 1979; Harrison et al. 2015a). The geological situation in the area of outcrop KPA758 is in parts stratigraphically unclear. Deposits of the Christopher and Kanguk Formations, and a small outcrop assigned to the Isachsen Formation, are mapped (Miall 1979). The latter represents a 77 m thick section of interbedded sandstone and mudstone with some intercalated thin coal seams and siderite horizons. Recent palynological investigations of this outcrop suggest an Early Cretaceous age for these deposits (Galloway et al. 2020). The combustion metamorphosed rocks of outcrop KPA758 form an isolated outcrop surrounded by Quaternary loose sedimentary rocks, and thus their relationship to unmetamorphosed equivalents is obscured.

2.2.2. Able River (North Banks Island)

In the outcrop KPA763 situated near Able River (Fig. 6b), Cretaceous rocks of the Isachsen, Christopher, Hassel, and Kanguk Formations are exposed. In the east, they rest unconformably on Upper Devonian rocks, and in the west, they are overlain by sedimentary rocks of the Eureka Sound Group and sedimentary rocks of the Beaufort Formation, respectively (Fig. 1b; Miall 1979; Harrison et al. 2015a, 2015b).

The Kanguk Formation in this area is locally folded in a tens-of-meters scale and affected by several faults (Piepjohn et al. 2018).

Close to outcrop KPA763, the basal bituminous shale member of the Upper Cretaceous Kanguk Formation (Miall 1979) is exposed and consists mainly of black to gray mudstone with numerous thin tuff horizons. A basal, 3-m thick fine sand layer below the Kanguk Formation is probably related to the Hassel Formation (Galloway et al. 2020). In the center of the biggest unconsolidated clinker mound, a 1-m high pillar of paralava and heated and consolidated mudstones are exposed. The dark bituminous shales of the Kanguk Formation provided the fuel for the combustion metamorphism. Recent palynological investigations suggest a marine depositional setting and a Cretaceous age for these deposits (Galloway et al. 2020).

2.3. Mackenzie Delta

Three outcrops with paralava and clinker (YU32, YU34, and YU69) occur along the banks of the Aklak Creek, a small tributary to the Mackenzie River Delta (Fig. 7). Along the Aklak Creek, Paleogene sedimentary rocks of the Moose Channel and Reindeer Formations (Fig. 1b) rest on deformed Lower and Upper Cretaceous rocks. The Reindeer Formation consists of sandstone, conglomerate, shale, and coal seams (Norris 1981) and is the equivalent to the Eureka Sound Group in the Canadian Arctic Islands (cf. Harrison et al. 1999).

2.4. Summarizing description of host sedimentary rocks

The thermally unaffected host rocks of paralava and clinker were not sampled for this study. Therefore, we refer to the literature in the following. The host rocks comprise typically terrestrial Paleogene sedimentary rocks on Ellesmere Island and in the Mackenzie Delta, as well as shallow marine Cretaceous sedimentary rocks on Banks Island (Table 1).

All host rocks consist of sandstones, siltstones, and mudstones and therefore represent clastic sedimentary rocks having their origin in eroded Precambrian crystalline basement rocks, as well as Paleozoic and early Mesozoic sedimentary rocks (e.g., Hadlari et al. 2015). Mountjoy (1967) noted the close lithologic similarity of Cretaceous and Paleogene sedimentary rocks of the Mackenzie River area and Banks Island with the Eureka Sound Formation in other areas of the Arctic Islands, including Ellesmere Island. The Eureka Sound Formation has been since redefined as Eureka Sound Group (Miall 1986; Harrison et al. 1999) and comprises the Paleogene sedimentary rocks only. We use the analyses of Riediger (1985) from Eureka Sound Group sedimentary rocks of the Margaret Formation from Ellesmere Island as representative to illustrate the general lithological composition of the examined clastic sedimentary rocks from the Mackenzie area and Banks Island, complemented by own observations. Riediger’s analyses show widely varying grain sizes and mineralogical compositions in Eureka Sound Group rocks on Ellesmere Island. The dominate minerals are quartz, feldspars, mica, and minor amounts of heavy minerals like zircon, tourmaline, garnet, rutile, kyanite, sillimanite, and others. Rock fragments of older sedimentary, igneous, and metamorphic rocks are also present. Various clay minerals are present in changing amounts and are concentrated in mudstone layers. Additionally, the clastic sedimentary rocks contain siderite concretions that appear to be of early diagenetic origin. More rarely, pyrite concretions are present. Cements like calcium carbonate, and less frequently siderite, occur in otherwise mostly unconsolidated sedimentary rocks.

These clastic rocks are frequently interbedded with coal seams of low coal rank or organic-rich dark silt and mudstones (Miall 1986; Harrison et al. 1999) that likely served as fuel for the spontaneous combustion processes starting with contact to oxygen near or at the current land surface.

2.5. Field observations on combustion products

At the visited outcrops, clinker pieces appear to be the most abundant combustion product and are often scattered over larger outcrop surfaces or else forming small ridges, where the more weathering-resistant clinkers have preserved original sedimentary structures, and in the case of the Banks Island material, fossil remains (Figs. 8a–8f). In contrast, paralava occurs in much smaller volumes (i.e., usually much less than ∼half a cubic meter) and commonly forms only a few spots within the scattered clinker pieces or remnants of the sedimentary rocks hardened by the combustion process (Figs. 8a, 8d, and 8e). The clinkers are often of a typical brick-red color. Less frequently, pieces of buff to brown or creamy color are present. Paralava has a grayish to black appearance, and some pieces show smooth surfaces that may show a glassy luster (KPA763; Fig. 8d). These features indicate likely flowing paralava over short distances. Broken pieces of paralava mostly have an internal vesicular texture. Often the paralava encloses pieces of the surrounding clinker resulting in a breccia-like appearance (breccia CASE12_01, KPA763, YU069; Figs. 8a, 8d, and 8f). Sometimes the contacts between paralava and enclosed clinker pieces are blurred. Breccias are present, where the organic matter of former coal seams is all burnt and the roof rocks (clinkers) are collapsed, or in chimney structures formed by ascending hot gases and vesicular paralava that intrudes overlying burnt and unburnt sedimentary rocks (Cosca et al. 1989; Heffern and Coates 2004; Grapes et al. 2011). The paralava outcrops at sample sites CASE12_01, KPA763, and YU34 (Figs. 8a, 8d, and 8e), where breccias are also present, probably represent residual chimney structures.

All outcrops featuring clinker or clinker and paralava have a weathered appearance, sometimes with orange–red lichen on the rock surfaces (e.g., Fig. 8d), indicating that the actual combustion process took place sometimes in the past. Thus, no subtle neoformations of ephemeral minerals were expected to be preserved, due to post-combustion weathering and erosion of the clinker and paralava pieces.

Samples of combustion metamorphic rocks were collected from 10 outcrops located on southern Ellesmere Island, Banks Island, and in the Mackenzie Delta area (Fig. 1; Table 1). For chemical analyses, macroscopically homogeneous pieces of clinker and paralava were selected.

The whole-rock powders of 19 clinker and paralava samples were analyzed for major and trace element concentrations by X-ray fluorescence (XRF), X-ray diffraction (XRD), and by inductively coupled plasma-mass spectrometry (ICP-MS). Preparation of the whole-rock sample powders, XRF and XRD analyses were performed at BGR, Hannover; ICP-MS analyses were carried out by Activiation Laboratories Ltd. (Actlabs), Canada, using analytical package 4B2 Research (lithium metaborate/tetraborate fusion, dilution by acid, and analyses by ICP-MS).

For XRF analysis at BGR, powdered samples were analyzed using a PANalytical Axios spectrometer. Samples were prepared by mixing with lithium metaborate Spectroflux as a flux material (flux No. 100A, Alfa Aesar) and melting into glass beads. The beads were analyzed by wavelength-dispersive XRF. The loss on ignition (LOI) was determined by heating 1000 mg of the sample powder to 1030 °C for 10 min. The calibrations are validated by analysis of reference materials. Materials of known composition and 130 certified reference materials were used for the correction procedures.

XRD was used to determine the mineral content (non-quantitative) of all samples using a PANalytical MPD Pro Θ-Θ diffractometer (Cu-Kα radiation generated at 40 kV and 30 mA), equipped with a variable divergence slit (20 mm irradiated length), primary and secondary collar, and a Scientific X´Celerator detector (active length 0.59°). The samples were investigated as continuous scan at a step size of 0.0167° from 2° to 85° 2Θ with nominal time per step of 10 s.

Four paralava samples were selected for additional characterization of their texture and composition by scanning electron microscopy (SEM) at BGR. The Zeiss Sigma 300 VP FEG SEM was operated in low vacuum mode and samples were scanned without any carbon or gold sputtering/coating. For microchemical analysis, the following detectors were used: Bruker Xflash® 6/30 EDX detector, high-definition backscattered electron detector, secondary electron (SE) detector, variable pressure secondary electron (VPSE) detector, and an in-lens detector for detection of secondary and backscattered electrons.

One sample of thermally unaffected coal from the outcrop with the burning coal scree (CASE12_21) was petrographically studied and analyzed for random huminite reflectance at BGR (see Dolezych et al. (2019) for technical details). Total organic carbon (TOC) and total sulfur contents of the coal sample and of all clinker and paralava samples were determined by combustion with a LECO CS 230 carbon/sulfur analyzer at BGR.

Geochronex Analytical Services Ltd., Burlington, ON, Canada, performed the 40Ar/39Ar incremental heating dating (10 temperature steps) on selected whole-rock paralava samples. The samples were wrapped in Al foil and loaded in an alumina vial together with LP-6 biotite to be used as a flux monitor. The samples and flux monitors were irradiated in the nuclear reactor. Flux monitors were run and the flux gradients (J values) were calculated. The Ar isotope composition was measured in a Noblesse noble gas static mass spectrometer (NU Instrument Ltd.). The 1300 °C blank of 40Ar was <10−11 cc STP.

For comparison, two samples of siderite concretions, collected from thermally unaffected Paleogene sedimentary rocks on Ellesmere Island, were studied by thin-section microscopy and analyzed by XRD, XRF, and ICP-MS as described above.

4.1. Mineralogical composition of clinker and paralava

The results of whole-rock XRD analyses of clinker and paralava samples are summarized in Table 2. Due to the mineralogical composition, the paralava samples are subdivided into silicate paralava (six samples) and almost pure iron oxide paralava (three samples).

The dominant mineral assemblage of the studied clinker and silicate paralava samples comprises hematite + quartz ± feldspars. Most samples additionally contain typical high-temperature/low-pressure minerals as cordierite–sekaninaite solid solutions, cristobalite or tridymite (high-temperature/low-pressure polymorphs of silica), and sillimanite. Cristobalite is the dominant high-temperature silica phase, and only one sample (YU69) contains tridymite instead of cristobalite. Sillimanite is present in four clinker samples, mostly as traces. Sillimanite can also be present as a detrital accessory mineral in the sedimentary protolith (Riediger 1985). However, accessory minerals are mostly not detectable by XRD.

The presence of an amorphous phase, most likely glass, in some paralava and clinker samples is detected by a “hump” in the background of the diffractogram (Table 2). A glass phase is to be expected in paralava (and also in clinker if the temperature is high enough for partial melting). However, only samples with a considerable amount of unaltered glass can produce a typical “hump”.

Iron oxide paralava is dominated by hematite and can also contain magnetite, clinopyroxene, as well as traces of other minerals (quartz, feldspar, cristobalite, and siderite).

Due to the low Na concentrations in the combustion–metamorphic rocks (see below), the identified feldspars comprise most likely newly formed Ca-rich plagioclase (anorthite) and K-rich feldspar (sanidine and microcline); however a part can be residual. Clinopyroxene is locally present in clinker and paralava from southern Ellesmere Island (CASE12_22 and _28) as well as in an iron oxide paralava sample from Mackenzie Delta (YU34). It can (1) originate from a volcanogenic detrital component in the host sedimentary rocks and (or) (2) be newly formed in calcite-rich host sedimentary rocks during high-temperature contact-metamorphism (e.g., Owens 2000). Clinopyroxene was detected for example in thermally unaltered sandstone of the Paleocene Mount Lawson Formation close to the Split Lake clinker outcrop (Estrada et al. 2009; see also supplementary Fig. S1) supporting the first possibility. Calcite is present as minor phase in many samples, but is lacking in the clinopyroxene-bearing samples, supporting the second possibility. A clinker piece from site CASE12_22 contains thin, parallel, yellowish bands and lenses, which consist of clinopyroxene, minor quartz, feldspars, as well as traces of cristobalite and hematite (XRD analysis). Thus, these yellowish bands represent former carbonate layers metamorphosed to calc–silicates (Fig. 9b).

The minerals (detrital) quartz, muscovite–illite, siderite, and calcite are interpreted to reflect mainly remnants of the original composition of the host rock lithologies. Two clinker samples (CASE12_20_cl1 from the burning coal site and YU32_cl1) do not contain typical high-temperature minerals but instead traces of muscovite–illite and ±siderite, indicating combustion metamorphism at relatively low temperatures. Typically, no paralava was found at these sites.

Calcite and siderite likely reflect a former cement of the host sedimentary rocks and siderite is additionally present as diagenetic concretions in the Paleogene host sedimentary rocks. Gypsum only appears in samples from Banks Island, which are characterized by relatively high total sulfur concentrations (0.4–1.0 wt.% S). The sulfur can originate from sulfide minerals (e.g., pyrite) transformed into sulfate minerals and sulfur gases during the oxidation at high temperatures, afterwards transformed into anhydrite and gypsum during cooling and post-combustion weathering.

Whitish precipitates observed regularly on surfaces of various paralava and clinker samples were identified as an assemblage of silica polymorphs (quartz, cristobalite, and tridymite), mullite, as well as traces of feldspar and cordierite according to XRD analysis on a piece from site CASE12_28 (Fig. 9c). Thus, these precipitates result from the combustion process.

In summary, most clinker and silicate paralava samples are characterized by the mineral assemblage quartz + hematite ± feldspars ± cristobalite (or tridymite) ± cordierite–sekaninaite ± clinopyroxene (or calcite) ± sillimanite ± glass. The clinker and paralava samples from south Banks Island (KPA758) show a slightly different mineral assemblage of hematite + cristobalite (+ traces of quartz and gypsum, ± cordierite). The presence of typical high-temperature/low-pressure minerals in most samples clearly supports the combustion–metamorphic origin of the studied rocks, although a detrital origin of parts of such minerals (clinopyroxene and sillimanite) cannot be completely excluded. The mineralogical composition of the paralava samples is very different from such of volcanic rocks.

4.2. Whole-rock geochemical composition of clinker and paralava

The major and trace element composition of clinker and paralava samples is summarized in Table 3 and Table S1. For comparison between samples from the study areas on Ellesmere Island, Banks Island, and in the Mackenzie Delta, the major and trace element concentrations are normalized to upper continental crust (UCC) and to chondrite values in the case of the rare earth elements (REE), and graphically presented in Figs. 10, 11, 12, and 13.

The major oxide concentrations of the 10 clinker samples vary considerably, 42.2–72.0 wt.% SiO2; 12.7–24.4 wt.% Al2O3; 3.1–36.8 wt.% Fe2O3; 0.7–10.1 wt.% CaO; 0.3–3.6 wt.% MgO; 0.8–3.3 wt.% K2O; 0.6–3.8 wt.% TiO2; and 0.2–1.9 wt.% P2O5. Silicate paralava compositions are in the same range as the clinkers, 40.4–68.15 wt.% SiO2; 11.9–21.5 wt.% Al2O3; 5.3–32.2 wt.% Fe2O3; 0.9–8.4 wt.% CaO; 0.8–3.7 wt.% MgO; 1.6–3.2 wt.% K2O; 0.5–2.4 wt.% TiO2; and 0.2–1.3 wt.% P2O5. Na2O concentrations are <1 wt.% in all clinker and paralava samples.

The three iron oxide paralava samples are characterized by their high total Fe2O3 contents (74–95.3 wt.%), and only minor contents of SiO2 (up to 8 wt.%), Al2O3 (up to 4.2 wt.%), CaO (up to 7 wt.%), and MgO (up to 3 wt.%). Iron oxide paralavas have the highest MnO concentrations of the dataset (0.8–1.0 wt.%). Only one of the iron oxide paralava samples (YU34_pa1) has a high P2O5 content of 1.7 wt.%.

The LOI varies between 0.3 and 6.9 wt.%. Low values <1 wt.%, which are to be expected from combustion metamorphic rocks, are only found in paralava and clinker samples CASE12_22 and YU69 as well as iron oxide paralava samples YU34. Elevated LOI values in the other samples are related to the presence of carbonate minerals (calcite and siderite) and gypsum (see Table 2). Small amounts of additional mineral phases containing H2O, OH− or CO3 groups may have been formed by low-thermal alteration during cooling or post-combustion weathering. LOI-free calculations of the major element concentrations used for graphical presentations are given in Table S1.

The UCC-normalized concentrations of SiO2, Al2O3, MgO, and K2O are in most samples close to or below UCC (Fig. 10). CaO is enriched (>2× UCC) in CASE12_22_pa1, _cl1, and CASE12_28_cl1 of Ellesmere Island, corresponding to the presence of clinopyroxene in these samples. Compared to UCC, P2O5 is enriched in most samples. Na2O that is generally very low in all samples (<1 wt.%) is not considered in Fig. 10. The paralava–clinker pairs from the same outcrop show similar major element concentrations. Only total FeO, P2O5, and MnO tend to higher concentrations in the respective paralava sample (except the Banks Island samples). However, the sample with the highest P2O5 concentration (1.92 wt.%) is a thermally relatively little affected clinker sample CASE12_20_cl1. Samples from Ellesmere Island show very high values of total FeO, TiO2, P2O5, and sporadically MnO (except CASE12_22 samples and CASE12_01_cl1), whereas in samples from Banks Island, only the total FeO concentrations are very high. Despite the broad variation of the major element concentrations, all clinker and paralava samples plot in the diagram after Herron (1988) in the fields for shale, Fe–shale, and (to a smaller portion) graywacke (Fig. 11), corresponding to silty–sandy mudstones as protoliths.

Similarities and regional differences are also indicated by the trace element concentrations of silicate paralava and clinker samples (Table 3; Figs. 12 and 13).

The chondrite-normalized REE patterns of most clinker and silicate paralava samples parallelize the UCC pattern (Figs. 12a-12c). However, some samples from Ellesmere Island (BACHE-RED samples) have steeper, light-REE-enriched patterns similar to a volcanic ash layer in the Mount Lawson Formation (“Lawson ash 1”) shown for comparison, and sample CASE12_28_cl1 is enriched in all REE. These samples show additionally higher concentrations of Zr, Nb, Ta, Sr (±Cr, Co, Ni, Cu) than the other samples (Fig. 13). Most trace element concentrations of the iron oxide paralava samples, particularly the light REE, are low compared to clinker and silicate paralava samples (Figs. 12 and 13).

A group of redox-sensitive trace elements with unusually high concentrations (Mo 302–353 ppm, U 44–49 ppm, Ni 233–542 ppm, Tl 2.7–8.3 ppm, V 848–891 ppm, Zn 333–533 ppm, Co 63–99 ppm, and Cu 124–158 ppm) is present in clinker and paralava sample KPA758 from south Banks Island (Fig. 14). These enrichments relative to UCC are associated with high concentrations of Fe (∼30 wt.% total Fe2O3) and S (∼1 wt.% S) as well as As (14–27 ppm; 2.9–5.6× UCC) that is below the detection limit of ICP-MS in most other samples (Table S1).

In the sample KPA763 from Able River, northern Banks Island, only Mo (20–28 ppm) and U (18 ppm) show enrichments >6× UCC (Fig. 12). Some of the redox-sensitive elements also show high concentrations in sample CASE12_28_cl1 from Ellesmere Island, mainly Co (314 ppm), Ni (700 ppm), Cu (238 ppm), and V (438 ppm), but in contrast to the KPA758 samples, they are associated with higher concentrations of Cr (452 ppm) and REE (Table 3, Figs. 12a and 14).

In summary, silicate paralava and clinker samples from the same outcrop are geochemically similar with a tendency to higher concentrations of total Fe, P, and Mn in paralava (apart from the Banks Island samples).

4.3. Texture and matrix composition of paralava

Three different silicate paralava samples and one iron oxide paralava sample were examined by SEM.

Sample CASE12_01_pa is a silicate paralava with 54.2 wt.% SiO2. The molten matrix consists of Si, Al, and K in a ratio of 2:1:5 as main elements. Vesicular textures include relatively large bubbles with dimensions of up to about 1230 × 1558 µm and more elongated tubular bubbles in other parts of the same sample (Figs. 15a and 5d). Secondary minerals on the surfaces of the vesicles are likely iron oxides.

Sample CASE12_22_pa (cf. Fig. 15) is a silicate paralava with 60.7 wt.% SiO2. The molten matrix consists of Si, Al, and K in a ratio of 21:8:2.5 as main elements. Secondary minerals on glassy surfaces are mainly iron oxides. Vesicles are of relatively smaller size compared to the first sample. The magnification of one of the vesicles shows growth of fine needles of an unknown mineral (Fig. 15b).

Sample KPA763_pa is a silicate paralava with 62.7 wt.% SiO2. The molten matrix mainly consists of Si and Al in a ratio of 4:1. Secondary minerals are mainly iron oxides. The vesicles appear overall smaller compared to sample CASE12_01 with 54.2 wt.% SiO2 (Figs. 15c and 15d).

Sample YU34_pa is an iron oxide paralava that caused significant interferences partly hampering SEM recording due to its high iron content (95.3 wt.% total Fe2O3). The sample shows typical rounded textures, which are well known from metallurgy, resembling blast furnace slags (Figs. 15e and 15f).

As seen in the SEM pictures, the size and type of gas bubbles resulting from the combustion process and passing through the molten rock differ from sample to sample and might be influenced by the viscosity, and thus by the SiO2 content of the paralava. However, the size and shape of the vesicles vary strongly even within the same sample, whereas the SiO2 contents of the SEM-investigated silicate paralava samples differ only slightly. In contrast, the very fine-porous vesicular texture of the almost SiO2-free iron oxide paralava differs from the silicate paralava samples.

4.4. 40Ar/39Ar dating of paralava

To date the individual combustion events, six whole-rock paralava samples were analyzed by the 40Ar/39Ar incremental-heating method. The summarized results are shown in Table 4. The complete Ar–Ar data are available as supplement (Table S2).

All samples yielded complex, moderately to strongly disturbed age spectra with a wide scattering of the apparent ages calculated for the heating steps (e.g., between <1 and 60.4 Ma for sample KPA758). Thus, total fusion ages cannot represent geological meaningful ages for the combustion events in most cases. Inverse isochron ages could not be calculated. Most samples show a U-shaped spectrum with the middle heating steps yielding the youngest apparent ages, which are interpreted to approach the true age (e.g., Kelley 2002). They are often below the limit of quantization of the method (<1 Ma). These paralavas most likely formed very recently in the late Pleistocene or Holocene (Table 4).

Only the sample from Bache Peninsula (BACHE-RED_pa2) provided a plateau age of 6.06 ± 0.18 Ma (Fig. 16a). This late Miocene age (Messinian stage) indicates that the host sedimentary rocks of the paralava BACHE-RED_pa2, coal-bearing Eureka Sound Group sedimentary rocks of likely late Paleocene age (Sweet 2009), were exposed close to or at the late Miocene land surface, thus enabling their oxidative combustion and metamorphism.

Paralava sample CASE12_22 yielded a U-shaped age spectrum. The two youngest heating steps form a small “plateau” (41.5% of the 39Ar released) with a mean age of 1.4 ± 0.1 Ma (Fig. 16b). We interpret this early Pleistocene (Calabrian stage) age as the best estimation for the maximum age of combustion metamorphism of late Paleocene to early Eocene sedimentary rocks of the Eureka Sound Group at outcrop CASE12_22 near Strathcona Fiord.

The unsatisfactory Ar–Ar dating results are most likely the consequence of several factors like the overall heterogeneous composition of the host rocks, incomplete melting, mixing of paralava with unmolten clinker material on a microscopic scale, effects of weathering and devitrification of the glass matrix, and the very young age of the thermal events.

4.5. Coal petrography and composition

The thermally unaffected coal sample CASE12_21 was collected from the surface at the outcrop with burning coal scree on southern Ellesmere Island (Figs. 2 and 3). The coal is identified as sub-bituminous with a mean huminite (vitrinite) reflectance (Rm) of 0.44%. The sample contains 45.05 wt.% of TOC and 0.57 wt.% of total S (cf. Table S3).

The inertinite- and mineral-poor sub-bituminous coal is rich in huminite group macerals, especially ulminite and gelinite, as well as in liptodetrinite. Few pieces of sporinite–macerals (yellow fluorescence), cutinite–macerals, and traces of resinite were observed. The sample is characterized by numerous cracks, which are rarely filled with exudatinite–macerales. The cracks are an effect of weathering, suggesting in-situ oxidation processes in the coal (e.g., Kus et al. 2015; Dolezych et al. 2019).

4.6. Siderite concretions

To identify possible sedimentary protoliths involved in the formation of iron-rich paralava, two samples of siderite concretions from the Eureka Sound Group of Ellesmere Island were included in this study.

Sample SE219/08 was collected from a thermally unaffected sand layer within the Paleocene Mount Lawson Formation in the vicinity of the Split Lake clinker outcrop (cf. Estrada et al. 2009; Fig. 1). It represents a diagenetically cemented volcaniclastic lens. Siderite forms the fine-crystalline matrix of this highly magnetic, volcaniclastic rock. Ash-sized (mostly 0.1–0.3 mm), angular particles within the matrix are pumice, volcanic glass shards, mineral fragments of plagioclase, alkali feldspar, quartz, clinopyroxene, rarely muscovite, biotite, chlorite, and some fragments of volcanic rock and organic material. XRD analysis revealed siderite, quartz, plagioclase, alkali feldspar, clinopyroxene, and zeolite (resulting from alteration of volcanic glass). An outcrop image and photomicrographs of the sample are provided in Fig. S1.

Sample SE229/08 was collected from the Margaret Formation at Stenkul Fiord, a few hundred meters to the NE from the clinker outcrops (cf. Estrada et al. 2009; Fig. 1). It is a typical roundish, 10–15 cm long, diagenetically formed concretion only consisting of fine-crystalline siderite (partly opacitized), and some silty quartz grains (thin-section examination and XRD).

Due to the siderite content, both samples have a high LOI value (SE219/08: 21 wt.%; SE229/08: 29 wt.%). LOI-free calculations yield total Fe2O3 of 43 wt.% for SE219/08 and 85 wt.% for SE229/08 (Table S1). Such iron oxide concentrations would be expected after high-temperature metamorphism of these rocks. The other LOI-free calculated major element concentrations (>1 wt.%) are

  • in sample SE219/08: 37 wt.% SiO2, 1.4 wt.% TiO2, 6.3 wt.% Al2O3, 2.5 wt.% M gO, 5.6 wt.% CaO, 1.3 wt.% Na2O, and 1.2 wt.% K2O;

  • in sample SE229/08: 8.5 wt.% SiO2, 2.6 wt.% Al2O3, 1.2 wt.% MnO, 1.6 wt.% CaO.

In contrast to the very low trace element concentrations of sample SE229/08, sample SE219/08 has relatively high concentrations in Ba (1403 ppm), Zr (348 ppm), V (134 ppm), Cr (151 ppm), Ni (137 ppm), and Nb (63 ppm). The chondrite-normalized REE concentrations of SE219/08 show a steep pattern that parallelizes the volcanic Lawson ash 1 pattern reflecting the volcaniclastic input (Fig. 12d). The REE pattern of sample SE229/08 is typical for carbonate rocks (showing in average lower REE concentrations than clastic sedimentary rocks) and is parallel to iron oxide paralava sample CASE12_28_pa1, but stronger depleted.

5.1. Origin of variations in element concentrations in silicate paralava and clinker

Most of the studied clinker and silicate paralava samples contain the high-temperature/low-pressure minerals cordierite–sekaninaite, cristobalite (or tridymite), and sillimanite as well as a glass phase in varying combinations and amounts (Table 2).

Despite these mineralogical similarities between the combustion products in the Canadian Arctic, some regional and local particularities of the geochemical composition are manifested. The geochemical composition of paralava and clinker depends on the composition of the involved sedimentary rocks including the inorganic components of the organic matter. Inorganic components of the coal comprise (1) clastic materials (clay, sand, and silt), (2) dissolved and precipitated salts in the pore water, (3) inorganic elements and compounds associated with the organic compounds of the biogenic matter, and (4) minerals formed during diagenesis and coalification, such as concretions of carbonates or sulfides (e.g., Rösler 1991; Ward 2002). The organic matter is almost completely burnt during combustion metamorphism, as seen from the very low TOC values of paralava and clinker samples (Table 3). Other elements (S, Na, Cl, and F) are also removed at a large part as fluids or gas.

During combustion, the non-volatile inorganic components of the organic matter remain as coal ash, which becomes partly transferred into melts at increasing temperature. Low-rank coal can have high ash contents and high iron concentrations in the ash, e.g., ∼30 wt.% total Fe2O3 in the ash of lignite of the Second Miocene Horizon of Lusatia, eastern Germany (Schrön et al. 1989). Paralava is partly sourced from the inorganic components of the organic matter in contrast to clinker. This can explain some differences between paralava and clinker from the same outcrop, e.g., the observed tendency to stronger enrichments of total Fe, Mn, and P in paralava samples (Figs. 10 and 11). The most common phosphate mineral in coal is apatite (usually fluorapatite), but monazite, xenotime, and aluminophosphate minerals (such as crandallite, florencite, gorcexite, and goyazite) are also present in most coals (Finkelman et al. 2019). However, apatite or other P minerals were not detected by XRD in our paralava and clinker samples. Heavy mineral concentrates (optical microscopy) as well as SEM observations, including EDX showed no evidence for P-rich mineral phases. Thus, P is probably bounded in several mineral phases, which are present in accessory amounts below the detection limit of the XRD.

5.1.1. Influence of volcanism

The relatively limited geochemical variations between silicate paralava and clinker samples from the same outcrop are superimposed by distinct regional and local specific elemental concentrations. On Ellesmere Island, paralava and clinker samples follow a magmatic trend in the TiO2 versus Ni diagram interpreted to be caused by depositional volcanogenic input, whereas the samples from Banks Island and Mackenzie Delta show a flat trend with variable Ni concentrations and relatively constant low TiO2 contents (Fig. 17a). Apart from Ti and Ni, paralava and clinker in Paleocene host sedimentary rocks (the BACHE-RED samples, CASE12_28_cl1, and CASE12_20_cl1) show higher concentrations of P, light REE, Zr, Nb, Ta, and Sr (±Mn, Cr, Co, and Cu) than the samples CASE12_01 and _22 from the upper Paleocene to Eocene Margaret Formation. We suggest this difference is caused by the Paleocene sedimentary rocks that were more strongly influenced by active volcanism during the time of deposition than the upper Paleocene to Eocene sedimentary rocks. This is expressed by the presence of many altered volcanic ashfall layers as well as eroded and redeposited clastic volcanogenic materials. Remains of alkaline, mafic to felsic Paleocene volcanic rocks (61–58 Ma) with high contents of Ba, Sr, Zr, Nb, and light REE are known from northeast Ellesmere Island (Estrada et al. 2010). The chemical composition of the volcanic ash layers is variable, dependent on the type of volcanism and the physicochemical conditions during the deposition. Some volcanic ash layers are enriched in Ti, Cu, V, Cr, and Sc (e.g., horizon Lawson ash 2), while other layers (e.g., horizon Lawson ash 1) are richer in Zr, light REE, and Nb (Reinhardt et al. 2013). At the Split Lake site (Fig. 1), clinker and paralava as well as thermally unaltered sandstone of the Mount Lawson Formation are enriched in Ti, Ba, Sr, Zr, Nb, and ±Cr compared to clinker and silicate paralava of the Margaret Formation from the Stenkul Fiord site (Estrada et al. 2009). Extremely high enrichments of aluminophosphate minerals can be diagenetically formed, where volcanic ash falls into a coal swamp as in the Margaret Formation of southern Ellesmere Island (horizon Margaret ash 1 with ca. 20 wt.% P2O5; Reinhardt et al. 2013).

5.1.2. Influence of redox conditions

High concentrations of the trace elements Mo, U, Ni, Tl, V, Zn, Co, Cu, and As are present in the two clinker and paralava samples KPA758 from southern Banks Island (Table 3; Fig. 14). Additionally, these samples have relatively higher total Fe (∼30 wt.% Fe2O3) and total S contents (∼1 wt.% S) than other samples analyzed. Sulfur was originally likely higher in the protolith, but was reduced during combustion metamorphism by removal of sulfur oxide gases. The trace elements V, Zn, Ni, Mo, and U are typically enriched under oxygen-poor conditions in bituminous sedimentary rocks, such as black shales, due to processes, such as adsorption on organic matter and (or) formation of sulfides in the water column or shallow sediment (e.g., Brumsack 2006). The deposition of the host sediments of the clinker outcrops from Banks Island under reducing, anoxic conditions is indicated by low Th/U and high V/Cr ratios of these samples (Fig. 17b).

It can be assumed that the protolith of the clinker and paralava from south Banks Island was a bituminous, pyrite-rich, carbonate-poor (low Ca, Mg, and Mn) shale deposited in an oxygen-poor, probably euxinic environment. The protolith of the samples from north Banks Island (showing high concentrations mainly of Mo and U) was presumably deposited in a less oxygen-poor environment, probably in a marginal position of the basin (e.g., Schröder-Adams 2014). The geochemical results indicate that the sedimentary protolith of this site (Nelson Head) also corresponds to the basal bituminous shale member of the Upper Cretaceous Kanguk Formation. Reddish clinker outcrops within this member are present ca. 70 km farther north from outcrop KPA758 in valleys of the upper Sachs River (Miall 1979; Smith et al. 2016). Comparative geochemical analyses of the unaltered possible host rocks are advisable.

The Cretaceous is an important episode for black shale formation with global deposition of these organic-rich sedimentary rocks, especially during the early Aptian and during the Cenomanian–Turonian (e.g., Jenkyns 2010). For example, black shales in the Proto-Atlantic of the Cenomanian–Turonian oceanic anoxic event are rich in Cd, Mo, Tl, Re, U, V, Zn, As, Ba, Cu, and Ni (Brumsack 2006). On Bathurst Peninsula, 175 km to the SW from Nelson Head (samples KPA758; Figs. 1 and 6), comparable trace element enrichments occur in smoldering dark bituminous mudstones of the Santonian to late Campanian Smoking Hills Formation, an age-equivalent of the Kanguk Formation (Grasby et al. 2022). A geochemical comparison between the clinker and paralava samples from Banks Island and mudstones and clinkers of the Smoking Hills Formation is shown in Table 5. The thermally unaltered mudstones of the Smoking Hills Formation as well as the combustion-metamorphosed products of both regions are characterized by very low Th/U ratios (<2). Thus, the Th/U ratio is indicative for original anoxic conditions of the sedimentary protoliths even after these rocks have experienced high-temperature combustion metamorphism. The metal element concentrations of the samples from north Banks Island are in the range of the Smoking Hills Formation, whereas the samples from south Banks Island are richer in Fe, Mo, U, Co, and Ni.

5.1.3. Local combinations of various factors

Additional factors, such as redox conditions during deposition and the original content of organic matter in the sediment (important for redox-sensitive elements like Co, Cu, Ni, and V), as well as the elemental behavior during diagenesis and combustion metamorphism, can favor local high elemental concentrations as present in clinker sample CASE12_28_cl1 (Figs. 12a and 13a). The high Th/U and low V/Cr ratios of the clinker sample (Fig. 17b) suggest that conditions during the deposition of the protolith were not reducing.

In summary, the regional and local particularities in the element concentrations of combustion products are more conspicuous than the geochemical differences between silicate paralava and clinker from the same outcrop. Although the concentrations of most elements were assumedly more or less affected during combustion metamorphism, these regional and local particularities are interpreted to be inherited from the unmetamorphosed host sedimentary rocks and reflect primary stratigraphic features.

5.2. Origin of iron oxide paralava

A special feature of combustion metamorphism in the Canadian Arctic is the presence of almost pure iron oxide paralava, found at several sites on southern Ellesmere Island and in the Mackenzie Delta (this study; Piepjohn et al. 2007; Estrada et al. 2009). The iron oxide paralavas are probably a special case of “metacarbonate slags”, described as small rounded combustion–metamorphic bodies of various mineralogical composition formed by decarbonation of carbonate rocks (Ciesielczuk et al. 2015). We assume that the iron oxide paralava of the Canadian Arctic originates from diagenetic siderite concretions that are widespread in adjacent thermally unaltered Paleogene sedimentary rocks and coal seams.

Geochemical similarities between iron oxide paralava samples and siderite concretions, like samples SE219/08 and SE229/08, support their genetic relationship (Table S1; Figs. 12d and 13d). The elevated MnO concentrations (0.78–1.02 wt.%) found in iron oxide paralava can be inherited from siderite concretions (e.g., 1.16 wt.% MnO in sample SE229/08). Siderite (FeCO3) is decarbonized during heating and decomposed to iron oxides and carbon oxide gases. Insights into this process are given by heating experiments on natural, impure siderite samples from coal mining dumps, imitating coal-waste fire conditions in a thermal chamber coupled with XRD (Ciesielczuk et al. 2015; Kruszewski and Ciesielczuk 2020). Siderite disappeared from the XRD spectra between ∼480 and ∼560 °C (Kruszewski and Ciesielczuk 2020). Magnetite (Fe3O4) crystallized from ∼400 °C and its content increased up to a saturation at around 660–680 °C or 940–1040 °C depending on the composition of the samples (Kruszewski and Ciesielczuk 2020). At the final temperature of 1200 °C after progressive heating, the samples were not yet completely molten and contained a mineral assemblage of mainly magnetite, olivine-group minerals, ± magnesioferrite (MgFe2O4), ± maghemite (γ-Fe2O3), ± clinopyroxene. The cooled samples (∼30 °C) mainly comprised magnetite (±maghemite, magnesioferrite, and wüstite), olivine, and ±quartz (Ciesielczuk et al. 2015; Kruszewski and Ciesielczuk 2020). These heating experiments show (1) siderite concretions are adequate protoliths for the formation of iron oxide combustion products and (2) the formation of the iron oxide phases takes place at lower temperatures than the onset of melting of the associated silicate host rocks.

The ongoing heating during natural combustion cannot considerably affect the newly formed iron oxide mineral assemblage (magnetite and hematite), but clastic sedimentary rocks within (or surrounding) the original siderite concretion can be molten and transformed into small amounts of glass matrix that are of extremely different chemical composition from the whole-rock iron oxide paralava. Electron microprobe analysis on glass matrix of an iron oxide paralava sample from Stenkul Fiord yielded 53 wt.% SiO2, 27 wt.% Al2O3, 15 wt.% CaO, and only 9 wt.% total FeO (Estrada et al. 2009). In the outcrops studied in this work, the iron oxide paralava samples (CASE12_28_pa1 and YU34_pa1, _pa3) are associated with clinkers, which bear glass and high-temperature minerals (Table 2). During subsurface combustion, the iron oxide nodules can partially or completely melt at very high temperature (>1200 °C). These melts can migrate into overlying clinker due to their low viscosity, whereas the related silicate melts with higher viscosity stay in the depth and are not exposed in the outcrop.

5.3. Geological and climatic conditions for the initiation of spontaneous combustion

5.3.1. Information from the active coal scree fire

A scree slope with continuous supply of fresh pieces of coal like the one described before (Figs. 2 and 3), is an ideal situation to start the process of spontaneous combustion that begins with the low-temperature (<100 °C) oxidation of the coal pieces, by their rapid exposure to the atmosphere (Grapes 2010). Once the ignition temperature of the coal is reached, combustion begins. Boundary conditions include (1) particle size and especially the surface area of the coal pieces, (2) the coal rank, (3) its heat capacity, and (4) the heat of the reaction. The oxygen content of the coal, the coal’s moisture content, humidity of the surrounding atmosphere, and especially the existence of pyrite, are additional factors (e.g., Nelson and Chen 2007; Deng et al. 2015). The low, sub-bituminous coal rank and the high amount of reactive huminite-group macerals of coal sample CASE12_21 favor self-heating. Low coal ranks are generally characteristic for the Paleogene coals (mostly lignites) of Ellesmere Island (Table S3).

The observed fire seems to be an annual event caused by the fresh break-off from the canyon wall supplying continuously unweathered coal pieces to the aerated scree, which is situated in a protected and sunny location favoring auto-ignition of the coal pieces.

There is only limited supply of coal pieces from the thawing canon wall, and it will be stopped by freezing winter conditions. In deeper parts of the scree, coal appears to have burnt completely, and it is possible that the presence of permafrost may hinder or halt the auto-combustion process.

The fire might have been also favored by the long-lasting exceptional warm and dry conditions during the summer of 2011. Thus, climate change with warmer summers becoming more and more frequent in high latitudes may further enhance the frequency and severity of spontaneous combustion there.

In general, the observed burning coal scree illustrates the process of clinker formation on the current land surface. No paralava was observed at that locality, and the clinkers do not contain high-temperature mineral phases, which seems to be caused by the dispersed nature of the fire.

5.3.2. Information from fossil combustion outcrops

On Ellesmere Island, combustion events are recorded from the late Miocene to the present. So far, three periods of combustion metamorphism are identified by 40Ar/39Ar incremental heating dating on whole-rock paralava, (1) 6.1 ± 0.2 Ma (late Miocene, Messinian stage) on Bache Peninsula, Ellesmere Island (this work); (2) 3.3 ± 0.5 Ma (Pliocene, Piacenzian stage) south of Stenkul Fiord, Ellesmere Island (Estrada et al. 2009); and (3) since ca. 1.4 Ma (Lower Pleistocene, Calabrian stage) to present day (this work).

The oldest event identified so far took place on Bache Peninsula at ca. 6 Ma (sample BACHE-RED; Figs. 4 and 5). A major glacio-eustatic drop in sea-level in the order of several tens of meters occurred during the late Miocene (e.g., Miller et al. 2005) and is discussed as one trigger for the Messinian salinity crisis in the Mediterranean Sea lasting from 5.96 to 5.33 Ma (Jiménez-Moreno et al. 2013). Generally, a lowered sea level in the range of tens of meters results in increased fluvial incision and augmented backward erosion onshore. This erosional process appears to have reached the unconsolidated and tectonically deformed Eureka Sound Group sedimentary rocks preserved in the tectonic graben on Bache Peninsula, during the late Miocene. The valley leading to Bartlett Bay (Fig. 4), being still the lowest topographic point today, formed the connection to the Miocene coastline, which was then situated offshore below modern sea level. As the preserved paralava and its related clinker represent a subtle remnant of the late Miocene land surface, they indicate erosional pre-shaping of the modern land surface at this time. The modern drainage pattern is still oriented approximately along the west–east-oriented graben axis and leads into Bartlett Bay (Nares Strait) towards the east.

Sedimentary rocks of Miocene age are lacking onshore over wide areas in the Canadian Arctic Archipelago. The next known occurrence is the Ballast Brook Formation on northern Banks Island that is thought to be of middle Miocene age (Fyles 1990; Fyles et al. 1994; Williams et al. 2008; Matthews et al. 2019). A wood piece of Miocene age from Cornwallis Island (cf. Fig. 1) is mentioned in Blanchette et al. (1991) and Mustoe (2018). Apart from these occurrences, offshore sedimentary rocks in the Beaufort Sea show a pronounced unconformity separating Miocene from Pliocene units. This hiatus is interpreted to be related to the global late Miocene sea-level lowstand (McNeil et al. 2001).

Combustion metamorphism during the Pliocene is indicated in the Stenkul Fiord area, Ellesmere Island (Fig. 1) by the Ar–Ar whole-rock age of a paralava sample dated at 3.3 ± 0.5 Ma (Estrada et al. 2009). Nearby sedimentary rocks of the early Pliocene (Fletcher et al. 2019) Beaufort Formation crop out between Vendom Fiord and Strathcona Fiord (Fig. 2). Here, repeated wildfires were documented by charcoal pieces (Mitchell et al. 2016; Fletcher et al. 2019). These authors also discuss lightning and higher temperatures during the Pliocene as the likely origin for wildfires. Nevertheless, this is no indispensable prerequisite for the spontaneous combustion of coal or other organic-rich sedimentary rocks getting in contact with the atmosphere, as described above for the recent coal scree fire.

At Strathcona Fiord, the coal seams of the Margaret Formation are likely the latest that were exposed by erosion at or close to the land surface (Fig. 2). Here, the likely youngest sites of combustion metamorphism with a Calabrian age (∼1.4 Ma; CASE12_22) or younger ages (<1 Ma; CASE12_01) are present. An earlier time of combustion can be expected for site CASE12_28 situated in the Mount Lawson Formation along the southwestern limb of the synclinal structure (Fig. 2).

The apparently recent ages (<1 Ma) of the samples from Banks Island (KPA758 and KPA763) and the Mackenzie Delta area (YU69) indicate possible fires during the Pleistocene or Holocene, probably during interglacial periods or after ice sheet retreat at the end of the Last Glacial Maximum that exposed the sedimentary units at or close to the land surface. On neighboring Bathurst Peninsula along the steep cliffs of Franklin Bay (Fig. 1), smoking “bocannes”, which are fed by bituminous Upper Cretaceous sedimentary rocks of the Smoking Hills Formation, assumedly stratigraphic equivalents to the bituminous rocks of the Kanguk Formation of south Banks Island, are active up to the present time (Mathews and Bustin 1984; Grasby et al. 2022).

Whether the combustion process in the geological past took place directly on the former land surface, i.e., similar to the burning scree fire described above, or with the burning material located in a depth of several meters below the former land surface, could not be determined. Presence of paralava, breccias and chimney structures, high-temperature mineral phases in paralava, and clinker (apart from clinker YU32) hint to subsurface conditions and fires with higher temperatures causing the melting of the host rock.

This paper examines paralava and clinker samples collected from various places of spontaneous combustion across the Canadian Arctic hundreds of kilometers apart and documents a site of active burning coal in present day high-Arctic climate conditions. The results are as follows:

  1. The site of burning coal scree exemplifies the ubiquitousness of the spontaneous combustion processes that begin in any place at or below the current land surface, where the required physicochemical conditions are reached. The main trigger to start the combustion process appears to be the exposure of sufficient fresh fuel sedimentary rock, such as coal or dark organic-rich sedimentary rock, to atmospheric oxygen. This may happen in diverse settings like along weathering cliffs, after landslides caused by thawing permafrost (e.g., Fraser and Reinhardt 2015), or erosional exposure of coal seams, etc., processes that currently occur more and more frequently in high latitudes due to modern climate warming.

  2. Despite different fuel for the spontaneous combustion process, like coal or bituminous sedimentary rocks and contrasting sedimentary environments of the host rocks, at first glance, the resulting paralava and clinker are closely similar.

  3. Differences occur mainly because of diagenetic features present in some of the host sedimentary rocks, e.g., siderite concretions that may result in formation of iron oxide “paralava” resembling blast furnace slag. Differing contents of SiO2 of the host material might have an influence on the viscosity of the molten material and the size and form of gas bubbles “frozen” in the solidified paralava.

  4. The documentation of regionally and locally different geochemical features of clinker and paralava, mainly their trace element fingerprints, can be useful for stratigraphic comparisons and for archaeological purposes to help to identify the origin of artefacts. Clinker and silicate paralava from Paleocene host sedimentary rocks of Ellesmere Island are characterized by high concentrations (relative to UCC and the other samples) of Ti, P, light REE, Zr, Nb, Ta, and Sr (±Mn, Cr, Co, Ni, and Cu) related to volcanogenic input. High concentrations of redox-sensitive trace elements as Mo, U, Ni, Tl, V, Zn, Co, Cu, and As, as well as very low Th/U ratios are typical for clinker and silicate paralava on Banks Island formed from Cretaceous bituminous shales deposited in an anoxic environment.

  5. Several reasons, including especially the very young ages of the combustion events, hamper regularly reliable 40Ar/39Ar age determination. However, if dating is successful, the ages can provide valuable information, for example, for the timing of tectonic processes, the exposure of sediment during interglacials, or in a more general way, the long-term landscape evolution over several millions of years. Therefore, it appears useful to add complementary dating approaches in the future like fission track analyses and other thermochronological methods.

  6. Another important feature of paralava and clinker is their enhanced weathering resistance compared to their mostly unconsolidated host sedimentary rocks, which may have been already removed widely by erosion. This makes them valuable witnesses of geological processes close to the land surface even after several million years. Additionally, the eye-catching, often brick-red color of clinkers, and shiny surfaces of dark paralava help to identify the spots of former combustion metamorphism in outcrops.

The authors thank all Canadian authorities, institutions, and local communities for granting the permits to work on Ellesmere Island, Banks Island, and in the Mackenzie Delta area. Many thanks to all local wildlife monitors during expeditions CASE 12 (Vendom Fiord in 2011), CASE 15 (Yukon North Slope/Mackenzie Delta in 2013), and CASE 16 (Flagler Bay in 2014), namely Peter Amarualik Jr., Simon Idlout, Pilipoosie Iqaluk, Deborah Iqaluk, Frank Dillon, Leonard Gordon, and Frank Paul. Samples from Banks Island were collected during an expedition lead by GSC together with Rod Smith (field party chief), Keith Dewing, and Andrew Durbano who was a field assistant. The Sachs Harbour Hunters and Trappers Committee and wildlife monitors John Lucas Sr., Trevor Lucas, and Kim Lucas are thanked for their assistance and participation in field activities. The Banks Island field work was funded under the GSC’s Geo-Mapping for Energy and Minerals Program, Western Arctic Project, Banks Island activity led by Rod Smith. Helicopter support was provided by pilot René Gysler and engineer Joseph Gourd of Great Slave Helicopters, Inuvik. This research was conducted under Northwest Territories Scientific Research License 15800, Inuvialuit Land Administration Right to Access Land # ILA16SN002, Parks Canada Research and Collection Permit # AUL-2016-21396, and approval of the Sachs Harbour HTC. Assistance of N. Perry, Parks Canada, Western Arctic Branch, was greatly appreciated. Access and use of Polar Bear Cabin on northern Banks Island was granted by Northwest Territories Department of Environment and Natural Resources. Tiffani Fraser and Tammy Allen (both YGS) helped collecting paralava in the Mackenzie Delta area and discussed features of smoldering Cretaceous dark shales in Yukon, as well as Loic Labrousse (Sorbonne Université, ISTeP, Paris). Thanks also to Maurice Colpron (YGS), who mainly organized the expedition to Yukon Northslope. The team of the Polar Continental Shelf Program (PCSP) provided perfect logistical support for these expeditions.

The authors also thank Stephen Rippington (Astute Geoscience Ltd, United Kingdom, formerly at CASP) and Harald Andruleit (BGR) for engaged discussion in the field and collection of samples on Ellesmere Island. Both contributed to an earlier draft version of the description of the burning coal scree.

Many thanks also to Martin Blumenberg (BGR) for providing organic geochemistry data of the coal sample and discussion of the results, Jolanta Kus (BGR) for performing the organic petrography of the coal sample, Andre Marx (BGR) for operating the SEM, Kristian Ufer (BGR) for XRD analyses, and Yakov Kapusta (Geochronex Analytical Services Ltd., Canada) for discussions on the Ar–Ar dating results.

The authors thank Keith Dewing (GSC Calgary) for providing an internal review. Constructive comments by journal reviewer Lotte Larsen, an anonymous reviewer, and associate editor Luke Beranek greatly improved an earlier version of the manuscript.

This paper represents NRCan contribution number/Numéro de contribution de RNCan 20220094.

All data are provided with the article and its supplementary materials.

Conceptualization: LR, SE

Funding acquisition: KP, JMG

Investigation: LR, SE, RD, NK, KP, JMG

Methodology: LR, SE, RD, NK

Project administration: KP

Visualization: LR, SE

Writing – original draft: LR, SE, RD, NK, KP, JMG

Writing – review & editing: LR, SE, RD, NK, KP, JMG

Fieldwork took place within BGR’s Arctic research project CASE and received no external funding. Expedition CASE 15 to the Yukon North Slope was co-funded by BGR, YGS, and the Université Pierre et Marie Curie (Paris, France). A cooperation between the Cambridge Arctic Shelf Programme (CASP) and CASE in 2011 enabled the participation of Stephen Rippington on expedition CASE 12 to Vendom Fiord. Fieldwork on Banks Island was supported by the GSC’s Geo-Mapping for Energy and Minerals Program, Western Arctic Project.

Supplementary data are available with the article at https://doi.org/10.1139/cjes-2022-0142.

This work is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.

Supplementary data