An investigation of a Devonian/Carboniferous Boundary section on the Bolivian Altiplano

Abstract The Devonian/Carboniferous Boundary (DCB) interval is associated with mass extinction, isotope excursions and a short glacial episode. This study investigates how boundary extinction and environmental change is expressed in the glacial high-palaeolatitudinal record of the Bolivian Altiplano (western Gondwana). A latest Devonian and early Carboniferous section has been investigated using sedimentology, palynology, total organic carbon and bulk δ13Corganic. The Colpacucho Formation is a Late Devonian shelfal–marine siliciclastic sequence. It is overlain in the study area by a unit of coarse sandstones and sandy diamictites, interpreted as glaciomarine. This distinctive glaciomarine unit is at least 7 km wide and 60–120 m thick with a variably incisive basal contact (<100 m). It is of very latest Famennian age and is a stratigraphic equivalent of proven glacigenic deposits across central South America. The offshore marine Kasa Formation overlies the glacigenic unit above a basal flooding surface. The DCB is 12 m above this flooding surface on the last occurrence of Retispora lepidophyta and significant palynological assemblage changes. This includes the loss of the Umbellasphaeridium saharicum phytoplankton bioprovince, endemic to Gondwana. Marine and terrestrial palynological extinctions are synchronous with a 2 ‰ positive carbon isotope excursion interpreted to be reflective of changes in organic matter delivery and preservation during an interval of environmental stress. These results inform wider debates on global environmental change and mass extinction at the DCB.


Introduction
End Devonian Mass Extinction (EDME) was a severe and distinct biotic crisis affecting terrestrial and marine ecosystems in the latest Famennian Stage (Fig. 1). It was coincident with a short glacial episode within the range of Retispora lepidophytaa cosmopolitan miospore of latest Famennian age (Maziane et al. 1999;Caputo et al. 2008;Isaacson et al. 2008;Lakin et al. 2016). Proven glacigenic deposits are described in central South America (Díaz-Martínez & Isaacson, 1994;Cunha et al. 2007;Vaz et al. 2007;Wicander et al. 2011;Caputo & Dos Santos, 2019) and the Appalachian Basin of North America (Brezinski et al. 2008(Brezinski et al. , 2010. These indicate a near-polar ice centre in western Gondwana and a temperate ice centre in the southern margin of Euramerica respectively (Fig. 2). EDME is also known as the Hangenberg Crisis in the Rhenish Massif standard succession. It was of 100-300 ka duration and has been divided into three main intervals . The lower crisis interval (or Hangenberg Black Shale event -HBS) is associated with high-totalorganic-carbon (TOC) black shales, positive carbon isotope excursions (PCIEs), and widespread marine anoxia, e.g. in Europe (Brand et al. 2004;Buggisch & Joachimski, 2006;Kaiser et al. 2013;Kumpan et al. 2013Kumpan et al. , 2014, China (Qie et al. 2015), Vietnam (Komatsu et al. 2014), Tibet (Liu et al. 2019) and North America (Saltzman, 2005;Cramer et al. 2008;Myrow et al. 2011Myrow et al. , 2013Over, 2020). The HBS is commonly interpreted as transgression, but it also contains regressive proxies and so could more likely represent increased terrigenous input onto carbonate shelves (Kaiser et al. 2011;Bábek et al. 2016). Extinction primarily affected marine organisms, such as ammonoids, trilobites and conodonts (Becker, 1992;Chlupac et al. 2000;Corradini et al. 2013;Kaiser et al. 2016). Recent reinvestigations of European and Vietnamese sections show a more complicated picture. Firstly, anoxic conditions sometimes persist into the lower Tournaisian (Paschall et al. 2019). And secondly, corresponding negative carbon isotope excursions (NCIEs) are recognized preceding the HBS in the upper praesulcata zone / lower costatus-kockeli interregnum zone and Devonian/Carboniferous Boundary (DCB) (Matyja et al. 2020;Pisarzowska & Racki, 2020;. The middle crisis interval is characterized by eustatic sea-level fall and deposition of the 'Hangenberg Sandstone' and equivalents (van Steenwinkel, 1993). Eustatic sea-level fall immediately below the DCB is supported by regressive facies and/or detrital indicators observed from diverse geological settings (see Kaiser et al. 2008;Weber et al. 2008;Kumpan et al. 2013Kumpan et al. , 2014Bábek et al. 2016;Carmichael et al. 2016). Kaiser et al. (2011) estimated c. 100 m of relative sea-level fall in Morocco, which is comparable to the <100 m of marine incision observed in central Europe and North America (van Steenwinkel, 1993;Brezinski et al. 2010).
Several factors have been proposed as causes of EDME, ranging from meteorite impacts, marine anoxia, global carbon cycle change, palaeoclimate and sea-level change, to magmatic activity ).

LN
Palynostratigraphy can tie Late Devonian global schemes and events into South America (e.g. Troth et al. 2011;, which allows for a reinvestigation of the DCB from a siliciclastic high-palaeolatitude area affected by glaciation. Our objectives are to: (1) revisit and describe a diamictite sequence from western Gondwana (Bolivian Altiplano); (2) interpret the palynological record; and (3) test whether positive isotope excursions in organic carbon can be recognized. These results can then be compared against the global record to inform debates regarding EDME and global environmental change at the DCB.
The study area is on the NE shore of Lake Titicaca, near the community of Chaguaya (Fig. 3b-c). There is an uninterrupted Devonian-Mississippian sequence that contains diamictites, the global index miospore Retispora lepidophyta and claystones suitable for palynological recovery (Díaz-Martínez, 1992;Díaz-Martínez et al. 1999;di Pasquo et al. 2015di Pasquo et al. 2015di Pasquo et al. 2015. These make it an ideal study area. Seventeen stratigraphic logs are presented. Log A is a road section called 'Villa Molino' and Logs B-Q were measured along an approximately strike-parallel, 7 km long topographic ridgeline (Fig. 3c). The stratigraphy is mapped from field observations, log sections and satellite imagery and compared to the regional lithostratigraphy of Díaz-Martínez (1996) and Grader et al. (2007) (Fig. 3c-d). The Cumaná and Siripaci Formations are not present. An equivalent unit to the Cumaná Formation has been identified ('Cumana Formation equivalent unit': CFEU); it is distinct, well exposed and crops out along the ridgeline ( Fig. 4a-b). The CFEU is an informal classification termed by this study. It is defined by its two distinct bounding surfaces: a lower erosive contact with apparent down-cut and an upper conformable contact into not well-exposed claystones. It can be correlated to the Cumaná Formation based on the presence of key palynological taxa (i.e. Retispora lepidophyta and Umbellisphaeridium saharicum). The top of the CFEU is used as a tie-point between log sections (Fig. 4c). The Siripaci Formation is presumed absent under the intra-Carboniferous unconformity (Fig. 3d).

Materials, methods and terminology
Sedimentary logs were measured in the field at 1:50 scale using a tape measure and with the aid of a Jacob's Staff and Abney level. The term 'diamictite' is used as a descriptive term that classifies poorly sorted sedimentary rocks with varied grain and clast sizes from clay to boulders (Flint et al. 1960a(Flint et al. , 1960b. The Moncrieff (1989) classification system is used to discriminate diamictites from other poorly sorted rocks. The classification of Evans et al. (2006) was used to interpret diamictite facies.  Whole-rock claystone samples were collected from outcrop at a shallow depth (<20 cm). All palynological processing was by standard methods (see Phipps & Playford, 1984), including HCl (37 %) and HF (60 %) followed by decant washing to neutral and sieving at 15 μm. This was followed by a brief short treatment in hot HCl (37 %) to remove neoformed fluorides. The samples were then rapidly diluted into˜200 ml of water and re-sieved before storing in a vial. Whole kerogen samples were not sieved at 15 μm and directly strewn after HF digestion.
Miospore schemes and index taxa discussed are from western Europe (see Clayton et al. 1977;Streel et al. 1987;Higgs et al. 1988;Maziane et al. 1999) and the Amazon Basin, Brazil (see Loboziak et al. 1986Loboziak et al. , 1999Loboziak et al. , 2000Loboziak et al. , 2005, 2003Loboziak & Melo, 2002;Playford & Melo, 2009. Biozones are defined on the first occurrences (FOs) of key miospore taxa (Fig. 1). The term 'phytoplankton' refers to the preserved cysts of acritarchs and prasinophytes. Particulate organic matter (POM) includes spore, phytoplankton and phytoclasts (i.e. plant debris) that exist as particulate fragments. Amorphous organic matter (AOM) is structureless under light microscopy and is likely formed in the water column and/ or sedimentary substrate via microbial activity (Pacton et al. 2011).
Palynological investigation was difficult due to a high degree of degradation typical of the area (see also Díaz-Martinez et al. 1999). Only the samples showing the best-preserved palynomorphs were counted to at least 200 specimens for statistical data, with all other samples used for presence/absence data only. Nevertheless, c. 75 % of counted specimens could not be identified to a species/generic level or even in open nomenclature. This has likely reduced the total taxon count and imparted a significant preservation bias as: (1) robust forms are likely to have preserved more than fragile ones, and (2) taxa with distinctive features are more readily identified over those with subtle defining characteristics easily obscured by degradation. As mitigation, certain specimens were grouped into larger categories for the relative abundances, such as genera (e.g. Umbellasphaeridium spp.) or sculpture characteristics (e.g. apiculate spores).
The carbon content in the samples for the TOC profiles was measured using a Carlo-Erba EA-1108 elemental analyser. Between 2 and 3 mg of both decarbonated and original sample were separately analysed with the machine calibrated using a low Total Carbon (TC) 'soil' standard (1.55 %). Between every 10 samples a check was made using the standard as an unknown.

4.a.1. Colpacucho Formation
The Colpacucho Formation is at least 560 m thick in Log I, but its basal contact was not observed (Figs 5a and 6). It is composed of claystones that contain siderite concretions and subordinate interbedded sandstones. Larger sandstone interbeds can reach up to 1 m thick, and are cross-bedded, laminated and/or massive with occasional channels. Where preserved (Log A), its uppermost 100 m coarsens upwards into massive, laminated, cross-stratified and/ or variably bioturbated sandstones. Bioturbation consists of Chondrites sp. with rare Skolithos sp. (Fig. 4d-f). The uppermost Colpacucho Formation is defined at the point at which the unit transitions from claystone-to sandstone-dominated. Where the sediment has been significantly bioturbated, there is a mottled texture. At 124-141 m height in Log A, the cross-stratified sandstones contain rare claystone rip-up clasts, gravel laminae and bidirectional cross-stratification (Fig. 4f).

4.a.2. 'Cumaná Formation equivalent unit' (CFEU)
The CFEU varies in thickness from 58 m at Log A to 140 m at Log I. Overall, it is composed of coarse sandstones, gravel and diamictites. Its basal contact is a single or occasionally stacked gravelstone and/or breccia-conglomerate overlying an erosional surface. Variable incision of c. 100 m into the underlying bioturbated sandstones of the Colpacucho Formation is inferred from the correlation of sections ( Fig. 6a-b). This basal surface marks a subtle yet defining change in sedimentary character, above which sandstones are more thickly bedded, coarser and non-bioturbated. The unit can be broadly split into: (1) a lower sandstone-dominated subunit; (2) a laterally and vertically discontinuous, poorly exposed interbedded sub-unit; and (3) an upper sub-unit of cross-laminated sandstones and diamictites (Fig. 6a).
The lower sub-unit is predominantly composed of two facies. The first comprises thickly bedded, cross-stratified and well-sorted medium to coarse grained sandstones. The second comprises matrix-supported and poorly sorted gravelly sandstones that contain gravel and conglomeratic laminae that occasionally overlie erosive scours (Fig. 4g). These facies broadly coarsen upwards.
The interbedded sub-unit consists of cross-laminated sandstones, claystones and laterally restricted poorly sorted purple siltstones and muddy sandstones. The latter are very similar to diamictite facies but lack the coarser sediment fraction and clasts. Hummocky and swaley cross-lamination was observed at 70 m in section Log H. It fines upwards into poorly exposed claystones in Log G (Fig. 6a).
The boundary between the lower and upper sub-units is typically marked by strongly cemented gravel beds that can be correlated between sections (Figs 4h-j and 6a). They have a common stratigraphic association along the ridgeline; they are exclusively found on the top surface of coarsening-upwards gravelly sandstone facies and are always overlain by diamictite (Fig. 7). They are poorly sorted, 5-20 cm thick and contain interspersed quartz gravel that is especially concentrated on the top surface. The top surfaces have a patchy, weathered exposure, and commonly host linear striations and grooves <1 cm deep ( Fig. 4h-j).
The upper sub-unit is predominantly composed of diamictites and sandstone facies (Fig. 6a). The diamictites are ≤10 m thick, stratified and matrix-supported clast-rich to clast-poor sandy diamictites. They have relatively straight contacts that are rarely sheared and mostly conformable (Fig. 4k). Exposure is recessive and tends to be obscured by modern soil profiles. Stratification is subtle, defined by colour banding, rare lamination and faint bedding, which can be non-planar (Fig. 4l). The matrix is micaceous, poorly sorted (from clay to gravel) and weathers a distinctive purple colour (Fig. 4m). Clasts are composed of quartz gravel/pebbles (randomly orientated) and arenite sandstone lithics ( Fig. 4m-n). A rare number of clasts show soft-sediment shearing (Fig. 4o). Clast size and content is highly variable and diamictites are in compositional continuity with muddy sandstones where clasts are absent (Fig. 6a). Larger sandstone lithic blocks up to 1 m in diameter were observed in Log I (Fig. 4p). Overturned and convolute laminae are common ( Fig. 8a-c). Cross-stratified sandstones are commonly interbedded with the diamictites; either as discrete stringers (<2 cm thick) or as metre-scale beds ( Fig. 8d-e). The sandstone facies become progressively more thickly bedded and ripple-marked with height (Figs 7 and 8f). Sheared contacts and convoluted and/or overturned stratification, including flame structures, were observed across the ridgeline but especially at Log A ( Fig. 8g-j). Immediately to the southeast of Log A there is a 22 m thick overturned sandstone that lies above the topmost diamictite bed and contained diamictite intraclasts ( Fig. 8k-l).

4.a.3. Kasa Formation
The Kasa Formation has a measured thickness of 240 m at Log A. The basal contact is conformable and laterally correlatable. The Kasa Formation is divided into a lower claystone-dominated unit and an upper interbedded unit (Fig. 5a).
The lower unit is c. 150 m thick. It is composed of claystones that contain siderite concretionary horizons and interbedded cross-laminated sandstones.
The upper unit is c. 90 m thick. Sandstones are cross-stratified, finely interbedded and with rare bioturbation on exposed bed surfaces. There are several thinly interbedded matrix-supported and clast-supported diamictite and conglomerate beds (<1m thick) that overlie sharp erosive contacts (Fig. 8m). Associated with these are sandstones with overturned laminae and sheaf folds ( Fig. 8no). The diamictite and conglomerate beds are typically lensoid with limited lateral extent. There is a preferred orientation in the clasts along a sub-horizontal fabric ( Fig. 8m -'Imbricated siderite clasts'). Clasts are well-rounded and primarily composed of siderite nodules and rarer quartz pebbles.

4.b. Palynology
Three palynological assemblage intervals (AIs) are identified. Changes in the miospore and phytoplankton fractions occur at the same stratigraphic levels and are discussed together. See Figures 9 and 10 for palynological plates of the taxa discussed, Table 1 for total assemblage abundances, and the Supplementary Material (available online at https://doi.org/10.1017/S0016756821000741) for the presence/absence data.
4.b.1. Assemblage interval 1: Retispora lepidophyta / Umbellasphaeridium spp. AI-1 is defined as the range of Retispora lepidophyta in the counted samples from sample 'I110' in the Colpacucho Formation to 'E3' in the lowermost Kasa Formation ( Fig. 5a-  The miospore fraction is relatively poorly preserved, difficult to fully speciate and of low diversity. It is characterized by the high relative abundance of R. lepidophyta, which comprises up to a third of the total miospore count. Morphologically simple, single-walled and non-apiculate miospore taxa are common (Punctatisporites sp., Calamospora sp. and Leiotriletes sp.) and comprise half of the total miospore fraction (Fig. 5b).
Age-diagnostic miospore taxa are rare and were observed in out-of-count presence/absence data only. However, AI-1 does contain the First Occurrences (FOs) of key taxa, such as Knoxisporites literatus, Indotriradites explanatus and Verrucosisporites nitidus (Fig. 5b). Long-lived Devonian and Early Carboniferous species, such as Densosporites annulatus, Emphanisporites rotatus and Retusotriletes incohatus, were also identified.
Age-diagnostic miospore taxa are extremely rare, and difficult to speciate with confidence. Only a single age-diagnostic species, Anapiculatisporites semicuspidatus, had its FO within AI-2.
The phytoplankton fraction is relatively impoverished compared to AI-1 and characterized by the high relative abundance of Gorgonisphaeridium spp., which accounts for 73 % of total identified phytoplankton. AI-2 is also associated with the long-lived Palaeozoic to Mesozoic phytoplankton genera Veryhachium

4.b.3. Assemblage interval 3: spore-dominated
The base of AI-3 is defined at the loss of the phytoplankton fraction (including Gorgonisphaeridium spp.) in sample 'A34' (Fig. 5b-d). It is spore-and phytoclast-dominated and ranges entirely within the upper Kasa Formation.
The miospore fraction is difficult to identify confidently due to poor preservation. However, Punctatisporites sp., Leiotriletes sp. and Calamospora sp. were common and comprise 49 % of the total miospore count. The rest are mostly single-walled apiculate genera such as Anapiculatisporites sp., Apiculatisporites sp., Apiculiretusispora spp., Claytonispora sp. and Raistrickia sp.
The following age-diagnostic miospore taxa had their FOs in AI-3 but are either extremely rare or limited to single occurrences. These are: Anapiculatisporites ampullaceus, Indotriradites dolianitii morphon, Indotriradites viriosus and Waltzispora lanzonii. Only a single specimen of W. lanzonii that conformed to its original type description was identified (see Daemon, 1974). Several unusual forms were observed with up to four poorly developed apical shoulders, which are comparable to those described by Playford & Melo (2010. Waltzispora sp. 1 is morphologically like W. lanzonii but lacks shoulder apiculation. Additional age-diagnostic species have their FOs within AI-3 but are known from the Late Devonian. These include: Aratrisporites saharaensis, Convolutispora major, Verrucosisporites congestus and Verrucosisporites depressus .
Phytoplankton are almost non-existent in AI-3. Only rare specimens of Quadrisporites sp. were observed in the counts. Specimens of other marine phytoplankton taxa were only sporadically observed in the presence/absence data.

4.b.4. Palynofacies
The palynofacies are dominated by terrestrially derived phytoclasts and spores, with only minor proportions of marine phytoplankton and amorphous organic matter (AOM) (Fig. 5e). Some broad trends are recognized. There is an upwards decrease in the relative abundance of the marine fraction ('phytoplankton' and 'AOM') in the uppermost Colpacucho Formation. The proportion of phytoclasts in the lower Kasa Formation (AI-2) is reduced. In contrast, there is a sudden increase in phytoclast content in the upper Kasa Formation (AI-3), coincident with the decrease in Terrestrial/ Marine (T/M) ratio in the counts.
4.c. TOC and δ 13 C organic 4.c.1. Colpacucho Formation Total organic carbon values typically vary between 0.3 and 1.5 %, but peak at 2.5 % in sample A-10, which is coincident with an increase in phytoclast content (Fig. 11).
Only those samples above the FO of Indotriradites explanatus (LE/LN zone) were processed for bulk δ 13 C org as this is the level in which global PCIEs have been observed ( Fig. 11; see also Kaiser et al. 2016).     There is a negative shift in bulk δ 13 C org values of c. 2 ‰ compared to the uppermost Colpacucho Formation. There is an upwards positive trend throughout the lower Kasa Formation (samples D1 to A27) of 1.6 ‰. Within this trend there is at least one PCIE, and potentially two. These are more clearly observed in Logs D and E where there are near-continuous runs of samples at 1 m intervals (Fig 11f).
The lower 2 ‰ PCIE is at 13-22 m above the base lower Kasa Formation and is coincident with a negative TOC excursion of 0.8 %. Its base (sample E3) contains the last counted occurrence of Retispora lepidophyta. Particulate organic matter is noticeably darker and more degraded in samples through the 2 ‰ PCIE compared to those above and below (Fig. 12). Immediately above the PCIE, both TOC and bulk δ 13 C org increase by c. 1 % and 1 ‰ respectively. This is accompanied by less degraded and more translucent POM and the lowest observation of AI-2 (Figs 5 and 12c).
Processed palynological recovery was sparse through the 12-22 m interval in Log E (2 ‰ PCIE), despite the average TOC values of 0.52 %. This suggests that the bulk of the organic residue in the PCIE samples was in the <15 μm fraction lost during standard palynological processing (i.e. washed through the 15 μm nylon mesh). To investigate further, those palynological samples between 13 and 22 m height in Log E were reprocessed using HF only and with the <15 μm fraction retained. The resulting residues contained AOM and neoformed fluorides (the latter a consequence of HF reacting with clay minerals and trace calcium in the sample). This means that AOM was preferentially lost during palynological processing.
The second PCIE (1 ‰) is at 68-72 m in Log E and is accompanied by a negative TOC excursion of 0.7 %. However, due to the break in section at 47-68 m height it is only partially sampled (Fig. 11).

4.c.3. Upper Kasa Formation
TOCs and bulk δ 13 C org increase markedly into the upper Kasa Formation to a maximum of 2.2 % and −23.9 ‰ respectively. This correlates with the total loss of the marine fraction and the increase in phytoclast content observed in AI-3 (Fig. 11). In the upper part, TOCs decrease to an average of 0.6 %.

5.a.1. Colpacucho Formation: open-marine
The Colpacucho Formation is an open-marine shelf environment based on the phytoplankton population observed in AI-1 and previous work (Díaz-Martínez, 1991). The claystone-dominated lower part is interpreted as offshore. The uppermost Colpacucho is interpreted as a prograding shoreface (Fig. 13a). The high degree of bioturbation in the shoreface sands suggests environmental conditions were either more favourable for biological life or allowed for greater preservation potential. An absence of body fossils, spreite and back-fill structures suggests that bioturbation was caused by soft-bodied organisms, such as annelid or nematode worms. Tidal processes are interpreted to explain claystone rip-up clasts, gravel laminae and bidirectional cross-stratification at the very top of the Colpacucho Formation at Log A (Figs 4f and 5a-b).
These results indicate a shallowing-upwards trend and transition from an offshore marine environment into a tidally influenced shoreface. This is reflected in the decrease in marine palynomorphs in the palynofacies counts (Fig. 5e).

5.a.2. Cumaná Formation Equivalent Unit (CFEU): glaciomarine and remobilization
The CFEU is interpreted as glacigenic based on the following criteria: (1) striations and grooves, (2) diamictites containing large lithic blocks, (3) sheared contacts and soft-sediment deformation, (4) sandstone stringers and channels interpreted as ice-bed separation and (5) the association between striations/grooves and diamictites, where the former are exclusively overlain by the latter. Furthermore, the palynology present (Section 4.b.1) confirm its equivalence to the regional and glacigenic Cumaná Formation (Díaz-Martínez & Isaacson, 1994;Díaz-Martínez et al. 1999) and global context of glacial diamictites (Lakin et al. 2016). Marine conditions are supported by the occurrence of the same phytoplankton assemblage in samples below, within, and above the unit (Fig. 5b).
The basal surface represents a major erosional event associated with c. 100 m of incision into the underlying shoreface sands (Fig. 13b). There is no evidence for direct ice contact at the base of the CFEU (no striations/diamictites, etc.). The simplest hypothesis is subaerial or submarine erosion following sea-level fall in the preceding Colpacucho Formation (Section 5.a.1).
The lower sub-unit is interpreted as a subaqueous proglacial fan system, analogues of which contain common coarse massive to cross-stratified sandstones (e.g. Hornung et al. 2007) (Fig. 13c). The coarsening-upwards trends are interpreted as the progradation of proglacial fans and ice advance. Rare, overturned beds (such as at c. 30 m height in Log G) are interpreted as soft-sediment deformation caused by rapid deposition and excess pore pressure (Talling et al. 2012). Grain-size segregation into gravelly and conglomeratic laminae may have taken place during flow separation (Carling, 1990).
The interbedded sub-unit is relatively distal, as shown by claystone facies (Fig. 13c). Poorly sorted muddy and silty sandstones are rare and are interpreted as thin debris flows. The hummockyswaley cross-stratification shows evidence of storms. The AI-1 palynology indicates that offshore marine conditions and terrestrial vegetation were unaffected by the advance of ice both regionally and globally. Interestingly, these facies do not contain dropstones, which contrasts with the dropstone-in-shale deposits typical of the Cumaná Formation (Díaz-Martínez & Isaacson, 1994). Assuming a glacigenic interpretation is correct, there are two possibilities: (1) there was a localized ice retreat in the study area; or (2) glaciers in the study area did not contain much lithified and/or exotic clast material.
The striated/grooved gravel beds, which typically mark the boundary between the lower and upper sub-units, are interpreted as subglacial ice traction onto soft sediment. Gravel would have been deposited via lodgement processes. Their consistent stratigraphic position at the top of coarsening-upwards gravel sandstones is interpreted to mark the point at which proglacial sands were overridden by the advancing ice sheet. No deformational structures were observed beneath the striations and grooves. Subglacial drainage may have lubricated the basal surface and inhibited the formation of glacio-tectonized structures. Alternatively, subglacial ductile shearing and/or ice-keel scouring can also form softsediment striations without deformation (Woodworth-Lynas & Dowdeswell, 1994;Le Heron et al. 2005;Vesely & Assine, 2014). However, no features typical of these mechanisms (i.e. stacked striated pavements or scour/berm structures) were identified and so an icetraction hypothesis is preferred. A non-glacial interpretation is also considered unlikely. The striations and grooves do not conform to typical definitions of tool marks and gutter casts, which are moulds or casts caused by erosion beneath a coarser unit into unconsolidated muddy sediment (see Myrow, 2003) The upper sub-unit is interpreted as subglacial (Fig. 13d). The diamictite facies mostly contain randomly orientated clasts and gravel, suggesting that lodgement processes were predominant. However, the presence of sheared sandstone clasts and overturned laminae suggests a mixture of both deformational and lodgement Sample 'E1' beneath PCIE Sample 'E9' peak PCIE Sample E13 -post PCIE Fig. 12. (Colour online) Palynological assemblages in samples E1, E9 and E13 that were processed through standard techniques. Samples E1 and E13 are beneath and above the 2 ‰ PCIE respectively. Sample E9 is at the peak of the 2 ‰ excursion and shows much more degraded and darkened POM compared with those samples E1 and E13. Scale bars are 500 μm.
processes. Lithic clasts were probably reworked from the underlying and interbedded sandstones due to their compositional similarity. Sandstone stringers and lenses within the diamictite beds are interpreted to have been deposited by basal melt-water films formed via ice-bed separation (see Piotrowski et al. 1999Piotrowski et al. , 2001. Overturned laminae in these sandstones are evidence for post-depositional remobilization. Subglacial drainage may explain the presence of larger, stratified sandstone beds (Evans et al. 2006). The 'ice-traction till' classification of Evans et al. (2006) incorporates a continuum of deformational and lodgement depositional features and stratified inter-diamictite sandstones that is comparable to the features described in the CFEU. The overturned 22 m unit near Log A is interpreted as localized slumping above a diamictite décollement. Convoluted laminations and flame structures at Log A are further evidence of remobilization and dewatering.

5.a.3. Kasa Formation: open-marine to pro-deltaic
The 239 m directly measured thickness at Log A is significantly less than the 600-1400 m reported regionally (Díaz-Martínez, 1991. Sedimentation rates could have been reduced here, or, more likely, it was eroded during the development of the overlying intra-Carboniferous unconformity. The lower unit is interpreted as offshore marine. The basal contact is straight, sharp and widely correlated, suggesting sudden retrogradation of the preceding CFEU as ice receded and the climate warmed (Fig. 13e).
The upper unit records the progradation of relatively proximal sandstone facies. This is a common regional trend and is interpreted as the progradation of deltaic systems onto a shelfal setting (Díaz-Martínez, 1991Díaz-Martínez & Isaacson, 1994;Díaz-Martínez et al. 1999;Isaacson et al. 2008). In this study, the lensoid conglomerates/diamictites, erosive contacts, overturned laminae, sheaf folds and siderite rip-up clasts are interpreted as reworked density flows and debrites in an inclined pro-delta or delta-front setting. Siderite rip-up clasts are likely to have been derived from claystone deposits up-dip.

Sea-level
In other studies, conglomerates and diamictites in the Kasa Formation are thought to have been triggered by proglacial outbursts during an early Carboniferous glaciation event (Díaz-Martínez & Isaacson, 1994;Isaacson et al. 2008). This is supported by evidence for two Early Carboniferous glacial events in western Gondwana in the Tournaisian and Viséan (Caputo et al. 2008;Lakin et al. 2016). Ice may have persisted above the CFEU updip from the study area. However, there are no independent ice indicators (i.e. dropstones, striations, etc.) observed to support this in Log A and so a reworking interpretation is preferred.
Retispora lepidophyta has a near-global extent in the latest Famennian and is an important index species owing to its short geologic range and distinctive morphology (Maziane et al. 2002). As the vertical range of AI-1 is concurrent with R. lepidophyta it is interpreted to represent the latest Famennian Stage (Devonian). An undifferentiated LE/LN zone is defined from the FO of Indotriradites explanatus to the LO of R. lepidophyta (Fig. 5b). Further biostratigraphic refinement is not possible due to the extremely rare occurrence of key taxa, including Verrucosisporites nitidus. This spore is also noted to be rare in the Amazon Basin  and has been reinterpreted in western Europe as an ecozone representing proximal environments (Prestianni et al. 2016). It may therefore not be a suitable marker species for age correlation. The Amazon Basin RLe/LVa zones could not be recognized due to the relative paucity and poor preservation of Vallatisporites sp. specimens.

5.b.2. Identifying the Devonian/Carboniferous Boundary
The extinction of R. lepidophyta is near-synchronous with the DCB as currently defined (see Higgs & Streel, 1993;Aretz et al. 2016). As such, the DCB is picked on the last counted occurrence of R. lepidophyta (Fig. 5a). Very rare occurrences of R. lepidophyta were observed above the picked DCB in the presence/absence data and are interpreted as reworked (see Supplementary Material available online at https://doi.org/10.1017/S0016756821000741). Reworked R. lepidophyta is not unusual and this has been described in Mississippian strata of the Amazonas Basin . The significant loss of phytoplankton taxonomic richness and assemblage overturns between AI-1 and AI-2 represents EDME as expressed in the high-palaeolatitude record (Fig. 5b-d). The miospore (terrestrial) and phytoplankton (marine) overturns occur synchronously and abruptly in the initial post-glacial marine transgression.
In the miospore fraction, the relative abundance increases of apiculate miospores immediately above the DCB suggest the ecological niche that R. lepidophyta occupied was almost immediately filled post-extinction. The long-lived and morphologically simple genera Punctatisporites sp., Leiotriletes sp. and Calamospora sp. were apparently unaffected.

5.b.3. Assemblage intervals 2 and 3 -Tournaisian
AI-2 is an impoverished assemblage dominated by long-lived spore and acritarch genera with simple morphologies, reflecting a post-EDME setting. AI-3, in contrast, is likely a depositional effect caused by the progradation of coarser terrigenous material in the upper Kasa Formation. This is supported by the palynofacies being almost entirely composed of terrestrially derived phytoclasts and spores. Those Late Devonian to Carboniferous spores whose FOs occur in AI-3 are either reworked or are more likely to be observed in the relatively proximal facies of the upper Kasa Formation.
AI-2 and AI-3 are undifferentiated Tournaisian Stage (Carboniferous) based on the FOs of the miospores Anapiculatisporites semicuspidatus, Indotriradites viriosus and single occurrence of Waltzispora lanzonii in A-23, A35 and A-33 respectively. These taxa are restricted to the Tournaisian AL-PD miospore zones in the Amazon Basin ( Fig. 1; . Although Playford and Melo (2010) discussed the possibility that W. lanzonii extends into the Viséan, they considered the few records in this stage to be more likely due to reworking. Díaz-Martínez et al. (1999) observed Dibolisporites distinctus (now Claytonispora distincta) and Raistrickia clavata in the Kasa Formation at the Log A road section. These species were not identified in this study, but their presence would suggest that mid-to late Tournaisian (PC/PD zones) sediments are present.

5.c.1. Potential controlling factors
The global correlation of PCIEs at or around the DCB in both organic and inorganic carbon implies a global mechanism. Widespread marine anoxia and organic carbon burial is the leading hypothesis . Bulk δ 13 C org data (as in this study) reflect the sum value of all organic matter from the rock sample, including both terrestrial (spores, phytoclasts) and marine (phytoplankton cysts, water-column derived AOM) sources. This is important as stratigraphic trends in bulk δ 13 C org can be influenced by processes that favour the delivery (and preservation) of marine or terrestrial organic matter (Davies et al. 2012;Könitzer et al. 2014). Furthermore, early diagenetic effects such as oxidation on the seafloor can influence the isotopic ratio of organic carbon (see e.g. McArthur et al. 1992). Different detrital organic fractions can also be preferentially depleted in carbon during early diagenesis (Benner et al. 1987).
This means there are two potential controls on the bulk δ 13 C org results described: (1) changes in the isotopic value of dissolved inorganic carbon in the oceans, resulting from increased global carbon burial, and/or (2) changes in organic delivery and preservation.
Palynofacies counts can provide constraint on the organic fractions present in the sample, including the overall proportion of marine vs terrestrial organic matter. In this study, however, AOM was absent on the palynological slides in most samples and yet observed in the un-sieved residues in some samples, meaning it was preferentially lost through the 15 μm mesh used during standard palynological processing (see Section 4.c). Therefore, the palynofacies counts in this study may not be representative of the bulk rock organic content.

5.c.2. Preliminary hypotheses
A PCIE at the DCB and the broad positive trend in δ 13 C org through the Kasa Formation are consistent with what is observed globally (see Saltzman, 2002;Saltzman & Thomas, 2012;Kaiser et al. 2016), which could suggest it is linked to increased global carbon burial. Considering, though, that the proportion of marine vs terrestrial organic material in bulk rock cannot be quantified (see above) it is premature to link the observed stratigraphic trends and PCIEs in this study solely to global mechanisms. The most positive bulk δ 13 C org values are in the uppermost Colpacucho and upper Kasa Formations, both of which are coarser (progradational) units with TOC maxima and higher phytoclast abundance. These results therefore compare well with Davies et al. (2012) who observed more positive bulk δ 13 C org in coarser sedimentary facies, which was interpreted as reflecting increased terrigenous input.
The characteristics of the organic matter through the 2 ‰ PCIE at the DCB (darker, sparser, more degraded) and the low TOC also suggest that changes in the delivery and preservation of organic matter are a significant controlling factor. It is possible that stress in the terrestrial and marine environments may have caused the reduction in POM (i.e. spores, phytoplankton cyst, plant debris) at the DCB ( Fig. 14a-b). This reduction is observed by the sparse POM in the >15 μm palynological fraction and supported by the negative TOC excursion. The remaining POM represents the residual degraded remnants of organic material in the sedimentary system. Assuming a steady flux of AOM, then a reduction in POM delivery would increase the relative proportion of AOM in the samples, resulting in a corresponding shift in bulk δ 13 C org , i.e. a 2 ‰ PCIE (Fig. 14b). As environmental conditions became less stressed, there was a return of phytoclast-and palynomorph-rich assemblages, resulting in a corresponding δ 13 C org negative shift (Fig. 14c). The palynological assemblage above the PCIE is the diminished Tournaisian AI-2, which shows that EDME occurred coincidentally with these changes in organic matter delivery and preservation. A similar mechanism is possible for the smaller and only partially sampled 1 ‰ PCIE.
The above hypothesis implies that AOM is of a lighter average δ 13 C org value than that of POM to cause a positive shift. Carbon fractionation between microbes is highly variable, by as much as 15 ‰ in modern marine phytoplankton (Hinga et al. 1994). It would be difficult to infer what types of microbe were responsible for the AOM production based on this study alone. However, a potential candidate could be green sulphur bacteria, which in one modern species has an average δ 13 C org-cell of −20.2 ‰ (Zyakun et al. 2009).

Alternative hypothesis of the Cumaná Formation Equivalent Unit
The CFEU is considered part of the Colpacucho Formation by Díaz-Martínez (1992, 1999 and Díaz-Martínez & Isaacson (1994). Based on its sedimentology (i.e. diamictite facies) and similar palynological assemblage (e.g. Retispora lepidophyta / Umbellasphaeridium spp.), the unit is considered equivalent to the Cumaná Formation in this study. Díaz-Martínez (1992) interpreted the diamictites and 22 m overturned beds at Log A road section to be non-glacigenic and the result of debris flows, mass remobilization and the sliding of slabs. Remobilization features complicate any sedimentological interpretation because it is difficult to distinguish between glacial diamictites and debrites from field observations alone (Eyles et al. 1985;Visser, 1994). A non-glacigenic interpretation is supported by the absence of exotic clasts and dropstone-in-shale facies in the CFEU, which contrasts with the typical Cumaná Formation (Díaz-Martínez & Isaacson, 1994). Furthermore, debrites are known from the overlying Kasa Formation, which supports a regional stratigraphic interpretation of a progradational clastic wedge on an unstable foreland basin situated NE of an active basin margin (Díaz-Martínez, 1991, 1996Sempere, 1995). There are numerous modern and ancient analogues of large-scale mass transport deposits in slope and deep-water settings along active margins (see Alves, 2015); however, this study area would be a unique example of such a system upon a shallow shelf.  A limitation of a purely large-scale remobilization hypothesis is that much of the supporting evidence for sliding slabs and blocks is limited to Log A (Díaz-Martínez, 1992). The CFEU is largely consistent across 7 km and is not observed to be laterally compartmentalized into sliding slabs and/or blocks (Fig. 6a). Also, there is no evidence for shearing or remobilization at the basal incision surface. The 22 m overturned beds at Log A are interpreted by this study to have slipped along a localized décollement surface above a diamictite bed, which would explain its uniqueness in the study area. The sandstone facies, which form the bulk of the topographic ridgeline, are largely depositional in texture, i.e. cross-stratified, laminated or with common ripple mark (Fig. 5a). Deformational features can be present in glacigenic environments, and the range of remobilized features observed are not atypical of an 'ice-traction till' interpretation (see Evans et al. 2006). Furthermore, the diamictites can be tracked laterally and have relatively straight contacts and so contrast with the remobilized conglomerates and diamictites in the overlying Kasa Formation, which are small-scale, lobate and/or associated with sheaf folds (see Sections 4.a.3 and 5.a.3).
A glacigenic interpretation is preferred based on the evidence described in this study. However, glacigenic vs remobilized hypotheses should be further tested by sedimentological investigation utilizing microfacies and petrographic analysis. Specifically, the <20 cm striated/grooved gravel beds and overlying diamictites could be sampled for orientated thin-section or magnetic fabric analysis to investigate any textural features indicative of subglacial processes (e.g. van der Meer, 2003). Additional work is also needed to understand the wider palaeogeographic relationship between Colpacucho and Cumaná Formations and the CFEU described in this study.

The DCB in the western Gondwana
The DCB interval has historically been difficult to identify and correlate in western Gondwana due to the absence/rarity of key fossil groups (conodonts, goniatites, etc.). The findings of this study, if replicated elsewhere in central South America, provide three additional criteria for correlating the DCB interval in western Gondwana and integrating it with the global record of EDME.
Firstly, the boundary lies immediately above diamictite deposits within the lowermost post-glacial sequence. The record of glaciation in the study area is therefore consistent with the wider record of glacial diamictites observed immediately below the DCB within the range of Retispora lepidophyta (Caputo et al. 2008;Isaacson et al. 2008;Lakin et al. 2016). Glaciation in the study area is the high-palaeolatitude equivalent of the regressive facies and proxies observed in EDME's lower and middle crisis intervals Kaiser et al. 2016). The inferred magnitude of incision (<100 m) beneath the CFEU is consistent with the 75-100 m of incision and sea-level fall observed immediately below the DCB in North America (Brezinski et al. 2010), Central Europe (van Steenwinkel, 1993 and Moroccan Anti-Atlas (Kaiser et al. 2011).
Secondly, the DCB is defined by the sudden loss of Retispora lepidophyta and Umbellasphaeridium saharicum phytoplankton bioprovince within the initial post-glacial sequence (Vavrdova & Isaacson, 1999). The additional increases in single-walled apiculate miospores and Gorgonisphaeridium spp. above the boundary may also provide additional biostratigraphic constraint in western Gondwana where index taxa (e.g. Vallatisporites vallatus, Verrucosisporites nitidus, Waltzispora lanzonii) are rare. These overturns conform with global reference sections where EDME's upper crisis interval is associated with palynological and marine extinctions during sea-level rise (Streel & Marshall, 2006;Kaiser et al. 2016). Furthermore, the early Tournaisian AI-2 in this study is comparable with a contemporaneous diminished palynological record in North America and Europe (Higgs & Streel, 1993;Higgs et al., 1988;Wicander & Playford, 2013).
Thirdly, this study identifies for the first time that the DCB in western Gondwana is coincident with TOC and δ 13 C org excursions, comparable to global reference sections ). The positive δ 13 C org excursion observed in this study area starts immediately at the loss of the R. lepidophyta, which correlated to the the upper EDME crisis interval and is synchronous with overturns in miospore and phytoplankton assemblages. However, the PCIE observed in this study more likely reflects changes in organic delivery and preservation during an interval of ecological stress rather than being a direct result of global organic carbon drawdown. Further work is needed to test the controls on the 2 ‰ DCB PCIE. This may include processing different maceral types (i.e. phytoclast, phytoplankton, AOM) separately for compound specific biomarkers which would identify the relative proportion of end-member values controlling bulk δ 13 C org.
6.3. On the cause of palynological extinctions at the DCB Anoxia has been suggested as a kill mechanism for marine extinctions in the lower EDME crisis interval . Paschall et al. (2019) identified sustained anoxia into the middle and upper EDME crisis intervals also. In this study, neither obvious 'black shale' facies nor a high TOC value was identified at the DCB, suggesting that marine anoxia was not a factor in the marine phytoplankton extinctions and loss of the U. saharicum bioprovince. However, anoxia cannot be discounted, as a low TOC in siliciclastic settings may not preclude anoxic conditions (e.g. Harding et al. 2011). Furthermore, even though AOM was qualitatively observed in the whole kerogen samples (i.e. those processed without the 15 μm mesh), there is no constraint on the AOM flux. However, marine anoxia alone cannot explain the coincidence of terrestrial extinctions observed in this study and in the upper crisis interval globally (plants, miospores, placoderms, etc). Due to the observed co-occurrence of marine (phytoplankton) and terrestrial (miospore) extinctions in this study being constrained to the initial post-glacial transgression, rapid climate change associated with the sudden retreat of global ice centres is proposed as the leading cause of extinction in the upper EDME crisis interval. However, additional elemental geochemistry, and Hg/TOC curves in Log E could test marine anoxia and/or magmatic activity as other potential causes of the observed extinctions.

Conclusions
The stratigraphy, palynology and chemostratigraphy of a Devonian/Carboniferous boundary section in western Gondwana have been described. A prograding latest Famennian shoreface (Colpacucho Formation) is incised and overlain by a glacigenic unit consisting of coarse sandstones, diamictites and striated/grooved gravel beds (Cumaná Formation Equivalent Unit). The CFEU is at least 7 km wide, 60-120 m thick, and overlies <100 m of incision. Its top surface is a sharp transition into offshore claystones of the lower Kasa Formation, an offshore marine unit recording progradation of regional deltaic systems. The DCB is identified at 12 m above the CFEU on the last occurrence of Retispora lepidophyta, with an increase in single-walled apiculate miospores, and loss of the Umbellasphaeridium saharicum phytoplankton bioprovince. The Tournaisian palynological assemblages are impoverished, and dominated by long-ranging genera with simple morphologies. Coincident with extinction at the DCB is a 2 ‰ positive excursion in bulk δ 13 C org . This is accompanied by a 0.8 % negative excursion in total organic carbon. It is proposed that environmental stress reduced the supply of particulate organic matter, which increased the relative proportion of amorphous organic matter in the whole-rock samples, thus, causing a shift in average bulk δ 13 C org . Glaciation in western Gondwana is time-equivalent to eustatic sea-level fall immediately below the DCB. Palynological extinctions occur stratigraphically above diamictites in the initial post-glacial sea-level rise. Terrestrial and marine palynological extinctions observed at the DCB (i.e. the upper EDME crisis interval) are likely related to rapid climate change associated with the sudden retreat of ice centres in western Gondwana and Euramerica.
Supplementary material. To view supplementary material for this article, please visit https://doi.org/10.1017/S0016756821000741