Role of vegetation in shaping Early Pennsylvanian braided rivers: Architecture of the Boss Point Formation, Atlantic Canada

Authors


Abstract

Vegetation is a major driver of fluvial dynamics in modern rivers, but few facies models incorporate its influence. This article partially fills that gap by documenting the stratigraphy, architecture and palaeobotany of the Lower Pennsylvanian Boss Point Formation of Atlantic Canada, which contains some of the Earth's earliest accumulations of large woody debris. Braided-fluvial systems occupied channel belts of varied scale within valleys several tens of metres deep and more than 12 km wide, and their deposits predominantly consist of sandy and gravelly bedforms with subordinate accretionary macroforms, high flow-strength sand sheets and rippled abandonment facies. Discrete accumulations of clastic detritus and woody debris are up to 6 m thick and constitute at least 18% of the in-channel deposits; they represent lags at the base of large and small channels, fills of minor channels and sandy macroforms that developed in central positions in the upper parts of channel fills. Sandstones with roots and other remnants of in situ vegetation demonstrate that vegetated islands were present, and the abundance of discrete channel fills suggests that the formation represents an anabranching, island-braided sandbed river, the earliest example documented to date. Although some sphenopsid and lycopsid remains are present, most woody fragments are derived from cordaitalean trees, and the evolution of this group late in the Mississippian is inferred to have exerted a significant influence on fluvial morphodynamic patterns. The formation records a landscape in which active channel belts alternated with well-drained floodplains colonized by dense, mature forests and local patches of pioneering, disturbance-tolerant vegetation. Lakes and poorly drained floodplains dominated by carbonate and organic deposition, respectively, were also present. A large supply of woody debris triggered channel blockage and avulsion, and active channel margins and islands within the channel belts were initially colonized by pioneer vegetation and subsequently stabilized by large trees. A similar alternation of stable and unstable conditions is observed in modern braided rivers actively influenced by vegetation.

Introduction

Interest in the interactions between vegetation and fluvial dynamics has grown considerably in the past decades (Gurnell et al., 2000) and the effects of vegetation are increasingly being incorporated into fluvial studies (Gibling et al., 2010; Fielding et al., 2011; Gibling & Davies, 2012). Recent advances have occurred in the study of landforms and sedimentary processes influenced by vegetation, ranging in scale from vegetation-induced sedimentary structures (Nakayama et al., 2002; Rygel et al., 2004; Rodrigues et al., 2006) to the large-scale planform of entire channel belts (Welber et al., 2012), and in fluvial styles associated with the evolution and modification in vegetation (Greb et al., 2006; Davies et al., 2011; Foreman et al., 2012).

Vegetation plays a significant role in the hydrology and sedimentology of present-day fluvial systems (Coulthard, 2005) and variously acts to stabilize or destabilize the landscape. Roots enhance bank strength (Tal & Paola, 2007; Davies & Gibling, 2011) and restrict lateral migration. Accumulations of woody debris may aid avulsion (Jones & Schumm, 1999; Slingerland & Smith, 2004) or, alternatively, promote island accretion and bar stabilization (Gurnell et al., 2001; Collins et al., 2012). The influence of such accumulations on fluvial systems has been described for many modern rivers (Gradziński et al., 2003; Brummer et al., 2006; Montgomery & Abbe, 2006; Bocchiola et al., 2008; Wohl et al., 2010). Although large plant debris is commonly preserved in ancient fluvial deposits, the potential impact of plant debris on the dynamics of ancient fluvial systems has received relatively little attention to date.

Fluvial sedimentology and geomorphology are relatively young sciences that matured long after anthropogenic modification in most river systems began (Montgomery & Piégay, 2003; Wohl, 2013). Indeed, such modifications have altered the dynamics of rivers to such an extent that facies models extrapolated from non-pristine rivers may be only partially relevant to ancient fluvial deposits. Interference with natural fluvial processes has resulted from the destruction of riparian and upland vegetation, the creation of artificial banks, removal of logs from channels to aid transportation (McCall, 1988) and land-use changes due to agriculture (Kingsford, 2000; Naiman et al., 2000; Brooks et al., 2003; Walter & Merritts, 2008). Even the Platte River of Nebraska (USA), classically considered as a reference for modern braided facies models (Smith, 1971; Miall, 1996), has been modified considerably by human activity since the 1850s (Joeckel & Henebry, 2008; Horn et al., 2012a). Currently, few major rivers remain sufficiently unaltered to allow reliable geomorphic assessments of the effects of vegetation; such rivers include the Tagliamento River of Italy (Tockner et al., 2003; Bertoldi et al., 2010), rivers of western Canada and Alaska (Desloges & Church, 1989; Brierley & Hickin, 1991; Froese et al., 2005) and some Australian rivers (Tooth et al., 2008; Jansen & Nanson, 2010).

It follows that the documentation of ancient in situ and transported vegetation plays a particularly important role in understanding the interaction between rivers and vegetation. Logs of the giant fungus Prototaxites are present in Lower Devonian fluvial deposits (Hueber, 2001), and scattered woody debris from the progymnosperm Archaeopteris is commonly observed in Upper Devonian and Mississippian strata (Cressler, 2006). However, the first large accumulations of woody debris occur near the Mississippian–Pennsylvanian boundary, when early gymnosperms spread through dryland tracts (Gibling et al., 2010; Davies et al., 2011; Davies & Gibling, 2013). Well-documented Pennsylvanian examples of log accumulations are preserved in the Pottsville Formation in Alabama, USA (Gastaldo & Degges, 2007), the South Bar Formation in Nova Scotia, Canada (Gibling et al., 2010), and the Blanche Brook formation of Newfoundland (Falcon-Lang & Bashforth, 2004, 2005; Bashforth, 2005). Other woody accumulations of similar age were reported in Poland (Gradziński et al., 1982) and the UK (Johnson, 1999; Falcon-Lang & Scott, 2000). These occurrences are relatively rare and the interpretation of fossil woody accumulations remains limited (Gradziński & Doktor, 1995; DiMichele & Falcon-Lang, 2012).

The Lower Pennsylvanian Boss Point Formation of Atlantic Canada provides wide coastal exposures of braided-fluvial deposits that contain a remarkable suite of woody debris accumulations, probably the largest number identified to date in Palaeozoic strata. Abundant woody debris was noted by Dawson (1868), Bell (1944), and later researchers Browne & Plint (1994) and Falcon-Lang & Scott (2000), but a comprehensive account of the relations between these accumulations and the depositional architecture is lacking. The present study aimed to fill this gap through implementation of the following specific objectives: (i) provision of a detailed account of the depositional architecture in the type locality; (ii) description of a series of new architectural elements composed of clastic sediment and woody debris; and (iii) setting of a palaeobotanical and sedimentological context for the woody accumulations. A facies model is presented that draws on information from modern vegetated rivers and which considers the effects of vegetation on fluvial dynamics. The significance of these dynamic interactions for the Palaeozoic greening of the continents is discussed.

The study provides, for the first time on the scale of an entire formation, a comprehensive account of the interactions between vegetation, especially log accumulations, and fluvial morphodynamics in the sedimentary record. Exceptional outcrop exposure has enabled the architecture of an ancient braided system to be linked explicitly to the role played by vegetation in controlling the sedimentation process, highlighting the profound influence of plant ecology on bar formation, channel filling and the formation of vegetated islands. Interactions between fluvial sedimentary processes and vegetation have been the subject of much attention in modern rivers (Nakayama et al., 2002; Rodrigues et al., 2006; Welber et al., 2012) but little of this growing understanding has hitherto been applied to the ancient record.

Geological Setting

The Lower Pennsylvanian Boss Point Formation (Yeadonian–Langsettian; Utting et al., 2010) of the Cumberland Basin (Fig. 1A) is well-exposed in coastal outcrops along the Bay of Fundy, Atlantic Canada. The Cumberland Basin lies between the Caledonia and Cobequid Highlands and is the westernmost depocentre of the regional Maritimes Basin (Fig. 1C), which was affected by strike-slip activity, rift extension and halokinesis (Waldron & Rygel, 2005; Gibling et al., 2008; Hibbard & Waldron, 2009; Waldron et al., 2013). The Cumberland Basin accumulated 7 to 9 km of Upper Devonian to lower Permian strata that are divided into several regional groups (Gibling et al., 2008; Fig. 1B). The Pennsylvanian succession, represented in its lowermost portion by the Boss Point Formation, is largely terrestrial, with coal-bearing intervals and a few marine incursions (Ryan et al., 1991; Grey et al., 2011).

Figure 1.

Geological outline of the study area. (A) Chronostratigraphic framework of the Maritimes Basin: Tourn. = Tournaisian; Steph. = Stephanian. (B) Stratigraphic scheme for the Cumberland Basin fill. The stratigraphic interval reported in Fig. 2 (Boss Point Formation) is shaded. (C) Geography of the Maritimes Basin (shaded area in eastern Canada) and geological sketch of Nova Scotia (Canada) and adjacent provinces (after Ryan & Boehner, 1994). The Cumberland Basin, positioned between the Caledonia and Cobequid Highlands, accumulated ca 8 km of strata during the Late Palaeozoic (Gibling et al., 2008). The white star indicates the studied type locality at Boss Point. The rose diagram indicates overall drainage towards the south-east for the Boss Point Formation (data in this study).

The type section of the Boss Point Formation, now part of the Joggins Fossil Cliffs UNESCO World Heritage Site, has received only modest attention since the original descriptions in the 19th Century (Logan, 1845; Dawson, 1868). At the type locality of Boss Point, the formation rests unconformably on the Mabou Group and is overlain conformably by the Little River Formation (Fig. 1B; Calder et al., 2005). Here, the Boss Point Formation is almost fully exposed in coastal cliffs and on the adjoining tidal platform, and comprises ca 1200 m of predominantly fine-grained to medium-grained sandstone (70%) and mudstone (25%), with subordinate conglomerate (5%). Gravel-sized clasts are chiefly intraformational calcite and siderite nodules and indurated mudstone, with minor extraformational (igneous and metamorphic) rocks. Thin limestones and coals are present locally. The type locality probably represents a central position in the basin, because the formation progressively coarsens towards the adjoining highlands with an increase in extraformational gravel.

Within the Cumberland Basin, the formation was mapped by Ryan & Boehner (1994). Browne & Plint (1994) divided the middle, well-exposed portion into 16 megacycles, which range in thickness from 15 to 100 m and can be traced with confidence for at least 12 km along strike (Plint & Browne, 1994; Fig. 2). Each megacycle is composed of a lower mudstone and an upper sandstone unit (Figs 3 and 4). Mudstone units are floored by non-erosive surfaces and represent floodplain and subordinate lacustrine deposits (Copeland, 1957; Lawson, 1962). Sandstone units are floored by deep irregular scours and represent braided-fluvial deposits (Browne & Plint, 1994) with distinctive upper flow regime structures locally present (Allen et al., 2011). Regional palaeocurrent analysis for the Boss Point Formation reveals complex patterns, with: (i) eastward to south-eastward drainage in the southern part of the basin; (ii) northward to north-eastward drainage in the northern part; and (iii) transverse drainage from the local highlands (van de Poll, 1966; Fralick, 1981; Gibling et al., 1992; Browne & Plint, 1994).

Figure 2.

Satellite coverage of the northern Bay of Fundy with the location of selected Boss Point exposures. Stratigraphic logs for the exposures are shown in Fig. 13.

Figure 3.

Stratigraphy of the lowermost 1000 m of the Boss Point Formation in its type area (see Fig. 2). The topmost 200 m (data not shown in the figure) are mostly represented by poorly exposed mudstone. BP-1 to BP-10 indicate the location of the wood accumulations analysed in detail. Megacycles are from Browne & Plint (1994). The sites measured in detail were selected because they host significant accumulations of woody debris. For symbols, see legend in Fig. 4.

Figure 4.

Legend for symbols used in the high-resolution outcrop tracings and stratigraphic columns.

Continent-scale restorations place the Maritimes Basin at palaeoequatorial latitudes (Roy & Morris, 1983; Scotese & McKerrow, 1990), with northward drift through the Carboniferous from ca 15°S to 2°N (Ziegler et al., 2002). Calder (1998) recognized a shift from semi-arid to more humid conditions across the Mississippian/Pennsylvanian boundary, and Allen et al. (2011) identified high-frequency oscillations and periods of strong seasonality in younger Boss Point strata, suggesting a shift from subhumid to semi-arid conditions. Accordingly, although a seasonal climate prevailed throughout deposition of the unit, fluctuations in the degree of seasonality of precipitation and sediment discharge apparently intensified during deposition of the upper parts of the succession.

Methods

The present data set was acquired via bed by bed logging, line-drawing of remote-sensing images and outcrop exposures, collection of palaeocurrent measurements, and compilation of a palaeobotanical data set. Sedimentological analyses and correlation of sections were discussed by Browne & Plint (1994), Plint & Browne (1994) and Allen et al. (2011); a summary of the facies associations described by these authors is provided in Table 1.

Table 1. Facies associations described for the Boss Point Formation in the Cumberland Basin. Braided sandstone units include the Braidplain facies association of Browne & Plint (1994) and the A1 to A2 facies associations of Allen et al. (2011). Floodplain units include the Lacustrine facies association of Plint & Browne (1994) and the B1 to B2 facies associations of Allen et al. (2011).
Facies AssociationDescriptionInterpretation
Braided-fluvial unitsTrough and planar cross-bedded and minor massive conglomerateGravel bed braided rives dominated by lower regime flows
Chiefly trough cross-bedded sandstone. Minor trough cross-bedded pebbly sandstone, ripple cross-laminated and planar cross-bedded sandstoneSand-bed braided river dominated by lower regime flows
Plane-parallel, sigmoidal and hummocky cross-bedded sandstoneSand-bed braided river dominated by flows transitional from lower to upper regime
Climbing-rippled and trough cross-bedded, sharp-based lenticular sandstone, locally fining-upward and with vegetation-induced sedimentary structuresSmall-scale channels. Levée and splay systems. Rapidly aggrading scour fills
Floodplain unitsThin sandstone and mudstone beds, coarsening-upward rippled sandstone, rare wave-rippled sandstoneCrevasse splay and mouth bar complexes, locally wave-reworked
Blocky mudstone with calcite nodulesVertic palaeosols
Claystone and coal. Abundant vegetation featuresOrganic hydromorphic palaeosols and swamps
Platy and papery mudstone, limestonesHydrologically open, freshwater lake, locally anoxic

The architecture of the formation (Table 2) was evaluated using: (i) tracing, classification and interpretation of stratigraphic surfaces visible on the tidal platform (Table 3), which mainly represents an along-strike section; and (ii) assessment of valley and channel fill dimensions by regional correlations, direct observations and cross-bed thickness estimates (Bridge & Tye, 2000; Leclair & Bridge, 2001). On the tidal platform, seaweed hampers sedimentological analysis. However, contrasts in erodibility are prominent in remote-sensing images, which comprise an aerial photograph set acquired at low tide with ca 1·5 m of pixel size, and an IKONOS coverage acquired at middle-low tide with a pixel size of 60 cm. The satellite coverage was taken from Google Earth and digitally stitched using Arcsoft Panorama Maker®.

Table 2. Architectural elements in the Boss Point Formation. Proportions are based on a 1200 m long stratigraphic cliff section measured at Boss Point. Elements DA, LA and LS are difficult to distinguish with certainty at some levels, and a combined proportion is indicated. Elements from Miall (1996), with the exception of elements RF and WD, which are defined here.
PositionElementDescriptionProportion (%)
In-ChannelGBGravel bars and bedforms. Scour-based and commonly flat-topped sheets. Plane-parallel stratified and high-angle cross-stratified2
SBSandy bedforms. Cross-stratification36
DA, LADownstream-accretion and lateral-accretion macroforms. Mounded bodies with large-scale inclined strata composed mainly of plane-parallel stratification6
LSLaminated sand sheets. Sheets of tabular plane-parallel stratification
RFRippled fills. Mounded channel fills, with internal tabular or inclined plane-parallel stratification12
WDWood debris accumulations. Heterogeneous cross-stratified to plane parallel-stratified bodies representing: (i) sheet-like lags resting on major erosional surfaces; (ii) scoop-shaped channel fills; and (iii) mounded macroform cores12
OverbankOFOverbank fines. Tabular bodies, internally plane-parallel stratified. Floodplain distributary channels as low-relief scoop-shaped bodies, erosionally based and flat-topped, with internal tabular plane-parallel stratification28
CSCrevasse splay. Bodies flat-based with convex-up top surfaces, internally plane-parallel stratified or gently convex-up with local centroclinal cross-stratification4
Table 3. Ranking of depositional and erosional surfaces recognized in the Boss Point Formation. On the basis of palaeoflow data, the extent is mainly along strike.
RankExtentReliefInterpretationProcessesEquivalent
Bridge(1993)Miall (1996)Holbrook (2001)
7>12 km, possibly 40 kmUp to 15 mBounding surfaces of megacycles; those below sandstone units represent regional valleysBasin-scale reorganization of depositional tracts7Valley fill
6>120 m, locally >500 mTypically 5 to 15 m, up to 40 mBasal surfaces of major channels, channel belts or minor valleysRegional avulsion or valley cuttingGroups of macroscale sets5 to 6Channel belt/nested valley
510 to 70 m, up to 120 m0·5 to 6 mBasal surfaces of minor channelsCutting of small channels and local avulsion4 (erosional elements)Channel fill
440 to 110 m0·5 to 5 mMacroform topsTermination of macroform deposition due to in-channel reorganization and local flow diversionMacroscale strata4 (accretionary elements)
3>10 m, up to 70 m0·5 to 2 mIntra-macroform surfacesErosion and accretion events during floodsMacroscale strata3Nested channel cuts
2Few metres, rarely decametresCentimetre to decimetreCoset-bounding surfacesMesoform migrationMicroscale to mesoscale cosets2
1Centimetre to metreCentimetreLamination and set boundariesMicroform and mesoform migrationMicroscale and mesoscale sets0 to 1

Prominent surfaces on the tidal platform were traced to the cliff exposures. The strike-parallel or dip-parallel nature of the exposures was inferred from 386 palaeoflow measurements of sets of trough and planar cross-beds and ripple cross-lamination at 23 sites. Palaeoflow measurements were also used to correct the apparent width of channel bodies in oblique sections. The architectural hierarchy is represented in terms of microforms (ripples), mesoforms (bedforms and unit bars), macroforms (compound bars), channels and valley fills (cf. Bridge, 1993; Miall, 1996; Holbrook, 2001). Taking a pragmatic approach to the methodology, the macroform classification (Table 2) includes both basic elements (GB, SB, DA, LA, LS, RS and CS) and element assemblages (OF) (cf. approach used in Ainsworth et al., 2011), and introduces an extension to the classic suite of elements to describe accumulations of woody debris.

The general sedimentary features of the formation were assessed using a comprehensive stratigraphic log that encompassed the whole 1200 m of exposure (Figs 3 and 4). In the central, well-exposed portion of the formation, a total of 450 m of strata with woody debris were described with more detail at different sites; these sections sample most of the fluvial intervals in the megacycles of Browne & Plint (1994). Detailed measured intervals were further described with outcrop line-drawings and palaeobotanical analyses (see below).

Depositional Architecture

Facies and architectural elements

Braided-fluvial units

Description

Sandstone-dominated units display complex internal architecture (Fig. 5) determined by the stacking and amalgamation of a diverse range of in-channel bedforms and macroforms. These units comprise gravel and sand sheets and lenses, inclined accretionary elements, laminated sand sheets (LS), abandonment fills and woody accumulations (Table 2). As used here, both basic and composite architectural elements are classified as in-channel or overbank, generated by sediment accumulation and reworked by channel migration (Jackson, 1975; Miall, 1996). The stratigraphy and architecture of megacyle 2, representative of the formation, is reported in Fig. 6, with additional outcrop interpretations in Figs 7 to 9. Tracings of the tidal platform represent areas enlarged from Figs 10 and 11 which, in turn, illustrate the geometry of the main bounding surfaces on the tidal platform.

Figure 5.

Archetypal diagram illustrating the Boss Point architecture in the type locality, intended to represent the ideal architecture and general scale of a megacycle. The architectural element abbreviations (Table 2) refer to overbank fines (OF), gravel (GB) and sand (SB) bars and bedforms, laminated sand sheets (LS), rippled stacks (RS), woody debris accumulations (WD), and lateral-accretionary (LA) and downstream-accretionary (DA) macroforms. Furthermore, (CS) indicates isolated crevasse splay bodies encased within overbank fines. The braided-fluvial and floodplain units contain the deposits grouped in the Braidplain and Lacustrine facies associations of Browne & Plint (1994), respectively. The major surfaces shown mark nested channel cuts (3), macroform tops (4), minor channels fills (5), channel belts or nested valleys (6) and regional valleys (7), respectively (cf. Holbrook, 2001). Symbols are shown in Fig. 4.

Figure 6.

Stratigraphy and architecture of the braided-fluvial unit in megacycle 2. The section location is illustrated in Fig. 10. The most prominent surfaces (see Table 3 for ranks) in the measured section have been extended from the cliff exposure to the along-strike tidal platform exposure. The main accumulations of woody debris are present between the 185 m and 195 m levels (thicker on the foreshore), where palaeoflow and palaeobotanical data have been collected. For symbols and architectural element abbreviations, see Fig. 4 and Table 2.

Figure 7.

Stratigraphy and outcrop tracing of part of megacycles 13 and 14, showing the interbedding of large woody debris-bearing macroforms with other strata. The WD element represents channel fills; the upper reaches 6 m in thickness, being the most prominent organic accumulation in the Boss Point Formation in the type locality. The accumulations are, in turn, capped by two well-developed DA elements marked by large-scale inclined surfaces (reference for horizontal bedding is reported on the left side of the field photograph). The viewpoint of Fig. 8 is also shown. For symbols and architectural element abbreviations, see Fig. 4 and Table 2. Person for scale is 1·6 m tall.

Figure 8.

Photograph and line-drawing of a macroform core accumulation cropping out in the tidal platform (part of megacycles 13 and 14). Meterage refers to the stratigraphic columns reported in Fig. 7. Stranded permineralized logs (shaded grey) constitute the core of a macroform formed by the accretion of plane parallel-bedded and trough cross-bedded sandstone. Palaeoflow indicators for trough cross-beds are sub-parallel to the long-axis orientation of the logs. Person for scale is 1·8 m tall.

Figure 9.

Stratigraphy and outcrop tracing of the lowermost portion of the braided-fluvial unit of megacycle 15. Macroform core and channel fill woody debris accumulations are interbedded with sandy bars and bedforms and subordinate laminated sand sheets. Similar to the macroform core accumulation of Fig. 7, palaeoflow indicators and log orientation (the latter collected only in the upper macroform core accumulation) are largely coincident. For symbols and architectural element abbreviations, see Fig. 4 and Table 2. Person for scale is 1·6 m tall.

Figure 10.

Outcrop map of the tidal platform near Boss Point, with palaeocurrent data. The satellite coverage was taken during middle to low tide. The exposure has been rotated, so that the strike of the Boss Point Formation is horizontal and the floodplain units have been shaded to highlight the architecture of the braided-fluvial units. Ranks of key surfaces are shown (see Table 2). On the basis of palaeoflow data from trough and planar cross-beds and ripple cross-lamination, the tidal platform is largely a strike section with some oblique intervals, whereas the cliff section is largely a dip section.

Figure 11.

Megacycle architecture and seventh-order surfaces of parts of the Boss Point Formation exposed at Boss Point, based on aerial photographs taken at low tide. The megacycles show repeated alternation between floodplain mudstone units and braided-fluvial units, the latter forming most of the outcrop. Enlarged areas (A) to (C) show some megacycle boundaries. Blue lines indicate non-erosive flat-lying surfaces related to flooding or regional avulsion of large braided systems. Red lines indicate irregular scour of incised regional valleys, locally removing the underlying floodplain mudstone unit to amalgamate with sandstone of the megacycle below.

Gravel bars and bedforms (GB) compose the lowermost portion of fining-upward successions (for example, Fig. 6, 165 to 175 m), rarely occupy narrow and steep-walled channels (for example, Fig. 6, 214 m) and more commonly form thin, erosionally based sheets that extend for a few tens of metres. Bedding is generally planar but includes gently inclined surfaces that dip parallel to palaeoflow. Planar and trough cross-bedded gravels predominate, ranging in size from granules to medium pebbles (Fig. 12A). The great majority of gravel material is calcium carbonate and siderite nodules, and extrabasinal clasts are subordinate (Fig. 3).

Figure 12.

Outcrop expression of the facies and architectural elements. (A) Metre-scale gravel bedform (GB), composed of trough cross-bedded gravel, resting on a basal channel scour. The element is interbedded with sandy bars and bedforms (SB) and hosts coalified plant debris (visible at upper left). Scale bar is 1 m long. (B) Sandy bars and bedforms (SB) composed of cross-bedded sandstone interbedded with laminated sand sheets (LS) compose the uppermost portion of a channel fill topped by overbank fines (OF). Scale bar is 1 m long. (C) Downstream-accretionary macroform (DA) enclosed between a woody debris accumulation (WD) and sandy bars bedforms (SB). Inclined beds of plane parallel-stratified sandstone indicate progradation of DA towards the left. Person for scale is ca 1·8 m tall. (D) Detail of a laminated sand sheet (LS) composed of plane parallel-laminated sandstone showing convex-up antidunal bedding. Scale bar is 50 cm long. (E) Rippled stack (RS) composed of sets of climbing ripple-laminated sandstone, related to channel abandonment. Pencil is 15 cm long. (F) Interbedded sandy bars and bedforms (SB) and woody debris accumulations (WD, woody debris indicated by arrows). The latter is floored by an erosional surface and represents the lag of a minor channel. The strip of flagging tape is 50 cm long. (G) Red beds composed of overbank fines (OF) with encasing minor floodplain channels (fc), the latter composed mainly of fine-grained rippled sandstone. Scale bar is 1 m long. (H) Drab overbank fines (OF) composed of interbedded organic shale and coal (dark-coloured units) and limestone (light-coloured units), representing the infill of shallow, anoxic floodplain lakes. Coin for scale (circled in white) is ca 2 cm in diameter.

Sandy bars and bedforms (SB) compose the bulk of the formation (Fig. 12B). Trough cross-beds are predominant and form stacked cosets up to 7 m thick cut by undulating erosion surfaces. Sets of ripple cross-lamination are locally associated with the cross-beds. Much less common are isolated planar cross-sets up to 1·5 m thick. The SB elements cover GB elements within major facies successions (Fig. 6, 173 m), fill discrete channel forms (Fig. 6, 195 to 200 m) and cap channel bodies (Fig. 6, 180 to 182 m).

Downstream-accretionary (DA) and lateral-accretionary (LA) macroforms are up to 5 m thick (Fig. 6, 216 m). Where preserved, top surfaces are locally mounded and inclined, dipping at up to 5° (downstream) or up to 10° (laterally), and the macroforms consist mainly of plane-laminated and ripple cross-laminated sandstone with planar, convex-up or concave-up bedding. Palaeocurrent patterns are varied, indicating drainage near-parallel to near-perpendicular to the inclined surfaces. In some cases, the relation between inclined bedding and palaeoflow is complex, with an example shown in Figs 7 and 12C: low-angle planar surfaces in the cliff dip downstream based on palaeoflow indicators, whereas inclined beds at the same level on the adjacent tidal platform appear to have accreted laterally (megacycle 15, Fig. 10).

Laminated sand sheets dominated by plane-laminated beds with current lineation appear in, and are common above, megacycle 3 (Fig. 12D). These sand sheets are composed mainly of plane-laminated and low-angle planar cross-laminated sandstone, with minor planar, convex-up beds that cap ripple cross-laminated sets (cf. Fielding, 2006). In places, they are difficult to distinguish from dipping, plane-bedded macroforms.

Homogeneous rippled stacks of sandstone (RS) are up to 6 m thick, with prominent climbing ripples (Fig. 12E). These stacks occur preferentially at the top of braided-fluvial units and commonly rest erosionally on LS units; they are uncommon in megacycles 2 and 3 (177 to 180 m level in Fig. 6) but become frequent from megacycle 6 upward. They are gently mounded where seen along strike (top of megacycle 6 in Fig. 10) and appear planar and gently inclined (<5°) along the dip direction. Based on their erosional bases and geometry, they appear to fill discrete channels, and their internal architecture accords with the convex-up strata sets of Huerta et al. (2011).

Woody accumulations (WD, abbreviation for Woody Debris) are composed of heterogeneous assemblages of clastic sediment and woody debris and form a suite of in-channel elements with diverse geometry and facies components (Fig. 12F); WD is abundant in the formation and is represented by: (i) lags resting on erosional surfaces; (ii) fills of discrete small channels; and (iii) macroform core accumulations. These occurrences are defined here as an extension of the suite of clastic architectural elements and are treated in detail below.

Interpretation

Gravel bars and bedforms (GB) represent stacked gravel dunes, with their palaeocurrents indicating prevalent downstream accretion. The abundance of calcite and siderite nodules indicates the predominantly intraformational nature of the gravels, which probably originated from erosion and reworking of floodplain palaeosols bordering the active channels. Sandy bars and bedforms (SB) represent dune trains, with the isolated planar cross-sets representing small linguoid bars (cf. Smith, 1971; Cant & Walker, 1978). Where SB elements cover GB elements, commonly filling discrete channels, the successions are interpreted as base to top assemblages of mixed gravel-sand bars.

Downstream-accretionary and lateral-accretionary macroforms are interpreted as large foreset bars, with their downstream or lateral orientation arbitrarily determined from the angle between accretion direction and palaeoflow. Angles exceeding 60° have been taken to indicate predominant lateral accretion (Miall, 1994). Local flow diversions and oblique orientations of some accretion surfaces with respect to the mean channel trend may result in components of downstream and lateral accretion in the same macroform (Cant & Walker, 1978; Rust, 1981).

Laminated sand sheets represent within-channel sand flats, with convex-up forms interpreted as antidunes and transitional barforms between dunes and plane beds. Rippled stacks reflect strongly aggradational events and correspond in some respects with the hollow (HO) elements of Cowan (1991) and Miall & Jones (2003), which these authors attributed largely to confluence scours. However, they typically cap LS deposits and are not observed to underlie SB deposits, militating against an interpretation as confluence scours. A new element (RS) is defined for these deposits in view of the unusual thickness of ripple cross-laminated sandstone.

Floodplain units

Description

Floodplain elements are poorly exposed and internal surfaces are difficult to discern. Predominant overbank fines (OF) are planar units of massive and laminated mudstone rich in vertical root traces, and with calcite and siderite nodules commonly concentrated in discrete horizons (Browne & Kingston, 1993). Minor sandstone bodies are less than 5 m thick and a few tens of metres wide, floored by irregular scours and topped by planar surfaces; they are composed of ripple cross-laminated and trough cross-bedded sandstone (Fig. 12G). Subordinate heterolithic sandstone and mudstone bodies (CS) are composed of plane-laminated and ripple cross-laminated beds, with an architecture of planar beds (Fig. 6, 162 to 165 m). The latter are exposed in the cliff but are virtually absent in tracings of the tidal platform, with the exception of a set of cryptic convex-up surfaces in megacyle 11 (Fig. 10). Examples occur mostly in the topmost 200 m of the formation, which is less fully exposed, and have been reported from other exposures (Cape Maringouin, Fig. 2; Plint & Browne, 1994).

Drab mudrock, carbonaceous shale, coal and organic-rich limestones are also present but subordinate (Fig. 12H). Coals and carbonates are the most extensive beds in the formation (possibly extending up to 40 km along strike: Browne & Plint, 1994) and abruptly overlie braided-fluvial units.

Interpretation

Overbank fines (OF) mainly represent well-drained floodplains with extensive carbonate-rich palaeosols. The minor sandstone bodies are interpreted as small floodplain channels (erosionally based sandstone bodies) and crevasse splays (plane parallel-bedded heterolithic bodies), respectively. Drab fines, coals and limestones represent lacustrine and poorly drained floodplains. Although minor components, their lateral extent suggests sharp and regional flooding events (cf. Ielpi, 2013).

Surface ranking

Hierarchical arrangements of bounding surfaces provide a convenient approach for reconstructing alluvial architecture (Allen, 1983; Friend, 1983; Bridge, 1993; Miall, 1996; Holbrook, 2001; Ielpi, 2012). In the Boss Point Formation, seven ranks were determined from the lateral extent and relief of the surfaces, as well as from relations with the bounded strata. Figure 10 is a comprehensive line-drawing of the Boss Point outcrop belt to show the complex spatial relations between differently ranked surfaces.

First-order surfaces include internal growth surfaces within mesoforms (for example, cross-set increments and plane-parallel laminae) and have a lateral extent of centimetres to a few metres. Second-order surfaces are mesoform boundaries (for example, cross-set and plane-parallel laminae set boundaries). Their lateral extent is usually a few metres, rarely reaching decametres, both along-strike and downdip. Third-order surfaces correspond to prominent surfaces within macroforms (for example, coset boundaries) and reflect both erosion and accretion events associated with flood pulses. The extent is typically more than 10 m and exceptionally up to 70 m (megacycle 10). Surfaces are mostly planar with a 5 to 15° inclination and rarely have a concave-up form. Where they are erosional, third-order surfaces can be compared to the nested channel cuts of Holbrook (2001).

Fourth-order surfaces are represented by preserved macroform tops and indicate termination of macroform deposition due to in-channel flow diversion. The lateral extent ranges from 40 to 110 m, and the surfaces are planar or convex-up. Where planar, the surfaces are flat-lying or gently dipping (commonly less than 5°, rarely up to 10°).

Fifth-order surfaces are scoop-shaped scours that represent minor channels. These surfaces have 0·5 to 6 m of relief and represent cross-cutting channels (Ghinassi, 2011) or local avulsion events (sensu Heller & Paola, 1996). Their lateral extent usually varies from 10 to 70 m and is up to 120 m where several scours are linked (top of megacycle 3). Fifth-order surfaces resemble the surfaces bounding the channel-fill elements of Holbrook (2001).

Sixth-order surfaces correspond to major erosional features and indicate regional avulsion (sensu Heller & Paola, 1996) or prolonged erosion that resulted in valley incision. A distinction between large deep channels and small valleys is difficult, if not semantic (Best & Ashworth, 1997; Fielding & Gibling, 2005; Strong & Paola, 2008). However, many sixth-order surfaces define bodies that could be classified as channel belts or nested valleys (sensu Holbrook, 2001); to simplify descriptions, they are termed channel belts here and include deposits laid down in the largest, major channels. The minimum lateral extent is 120 m, and the maximum lateral extent, not observable, is greater than 500 m (megacycle 10). The erosional relief, usually ranging between 5 m and 15 m, may reach up to 40 m over as little as 130 m of along-strike exposure, and some surfaces cut the entire braided-fluvial part of the megacycle.

Seventh-order surfaces constitute the boundaries of the braided-fluvial (erosional) and floodplain (depositional) units and represent major, regional changes in the Boss Point depositional system. Browne & Plint (1994) interpreted these surfaces as generated by tectonically mediated base-level changes at the scale of the entire basin. Although the extent and style of termination cannot be determined from individual outcrops, the latter authors traced individual surfaces confidently for 12 km (Fig. 13) and possibly for up to 40 km. In the Boss Point location, erosional seventh-order surfaces are irregular with up to 40 m of relief and may entirely cut out the underlying mudstone (Fig. 11). In contrast, depositional seventh-order surfaces appear flat-lying or convex-up, with up to 5 m of relief over 50 m of lateral extent (megacycle 6 in Fig. 10). Although neither lateral terminations of erosional seventh-order surfaces nor correlative interfluve palaeosols are observed, the significant erosional relief associated with the surfaces suggests an interpretation of regional valley boundaries (Fig. 11), comparable to those bounding the valley fills of Holbrook (2001).

Figure 13.

Along-strike correlation of megacycles in six exposures of the Boss Point Formation (Fig. 2) along the Bay of Fundy, redrawn after Browne & Plint (1994). Seventh-order surfaces that bound the megacycles are shown, indicating that individual megacycles extend for at least 12 km along strike, having tens of metres of basal incision.

Width and thickness of channel bodies and valley fills

Correlation of seventh-order surfaces over a 12 km transect (Fig. 13; Browne & Plint, 1994) shows that no braided-fluvial units are represented in their entirety. Ten braided-fluvial units are interpreted here as valley fills with minimum width ranging from ca 3·6 to 11 km and thickness from 8 to 60 m. The average minimum width : thickness ratio is 367 ± 253 (area C in Fig. 14).

Figure 14.

Width : thickness space for Boss Point Formation channels and valleys, plotted on a base 10 log to log scale. The geometry of 72 minor channels (fifth-order basal surfaces) with near-complete exposure and preservation is shown in (A), and the minimum dimensions of 10 regional valleys (seventh-order basal surfaces) in (C). The area of (B) indicates the estimated range of dimensions for major channels (sixth-order basal surfaces) based on statistical analysis of cross-bed thickness (Leclair & Bridge, 2001) and width : thickness ratios of the population shown in (A).

Sixth-order surfaces represent channel-belt bases (Table 3) and also are not fully exposed, making their original dimensions difficult to assess. In contrast, 72 minor channel bodies floored by fifth-order surfaces are exposed nearly fully (Fig. 10). Most of the bounding surfaces are truncated, so that width and thickness estimates represent minimum values, although probably close to the original dimensions. The preserved channel bodies range from 2 to 89 m in width, averaging 25 ± 17 m, and from 0·5 to 5·5 m in thickness, averaging 3·5 ± 1·9 m. Width : thickness ratios as preserved are 1·7 to 30, averaging 7·8 ± 5·2 (area A in Fig. 14).

In the cliff exposures, 25 near-fully preserved channel fills are represented by fining-upward successions that pass upward into mudstone plugs. Channel fills range from 1·7 to 10·2 m thick, averaging 5·5 ± 1·9 m. These are minimum estimates, since the mudstone tops are truncated. The fining-upward rhythms rest on fifth-order or sixth-order surfaces and variously represent the deposits of both minor channel bodies and channel belts.

A total of 124 cross-sets range from 10 to 150 cm in thickness, averaging 45 ± 30 cm. Bridge & Tye (2000) and Leclair & Bridge (2001) demonstrated that cross-set thickness scales with the original dune height. The standard deviation : mean thickness ratio of the 124 cross-sets is 0·66, within the suggested 0·88 ± 0·3 value of the applicability test (Bridge, 2003). Multiplying the mean set thickness by 2·9 (Leclair & Bridge, 2001) suggests that the mean dune height averaged 1·3 m. This value represents dunes from numerous channels. Applying flow depth : dune height average ratios that are estimated to vary from 6 to 10, the formative flow depth for channels of all scales would have been in the general range of 8 to 13 m.

The thickness and width of individual channel bodies provide an approximation of the original depth and width of the channels. In summary, minor channels on the tidal platform, bounded by fifth-order surfaces, have maximum thicknesses of 5 m and widths of 90 m (area A in Fig. 14), with width : thickness ratios typical of narrow and broad ribbons and narrow sheets (sensu Gibling, 2006). Fining-upward units in the cliff are generally thicker, averaging 5·5 m but including units more than 10 m thick, and some represent major channels. The statistical analysis of cross-beds suggests that the formation was laid down in channels in the range of 8 to 13 m deep. Thus, nested within the regional valleys, the Boss Point system comprised major and minor channels with a large range of dimensions, commonly 5 m and perhaps as much as 13 m deep. Applying a width : thickness ratio from 8 to 30 (average and upper end-member, respectively, for fifth-order channels), the major channels may have been in the order of 40 m to 400 m wide (area B in Fig. 14).

Woody Material

Analyses and general characteristics

The size distribution, preservation and taxonomy of plant remains in the Boss Point Formation were examined to differentiate between various types of woody accumulation. Based on the palaeoecology and growth habit of the plant groups, the provenance of the woody material and its potential impact on fluvial dynamics is also addressed. Almost all plant remains observed in the succession are parautochthonous or allochthonous (sensu Gastaldo et al., 1995) and include parts of uprooted tree trunks, canopy branches, bark fragments, transported roots and possibly petioles of large pinnate leaves. Vegetation in growth position is rarely preserved (Fig. 15A and B), but upright stumps with attached root systems, sediment-filled axes and traces of root networks (Fig. 15C and D) are present. In situ plant fossils are largely confined to floodplain units, although five occurrences of autochthonous axes were observed in braided-fluvial units.

Figure 15.

Examples of in situ vegetation preserved in the Boss Point Formation. Compass in (A), (C) and (D) is 10 cm long. (A) Coalified cordaitalean tree stump with partially preserved root system, in the floodplain mudstone unit of Megacycle 2. The stump tilting reflects differential compaction. (B) Standing, sediment-filled cordaitalean in channel sandstone near the base of braided-fluvial unit of megacycle 3. Scale bar is 50 cm long. A detailed stratigraphy and shore tracing of the larger outcrop is reported in Fig. 21. (C) Mottled red mudstone with roots associated with local reduction aureoles. (D) Detail of a reduction aureole around a root, with some coalified organic material preserved.

The composition and size of parautochthonous to allochthonous vegetation was quantified at 10 levels in the type section (Figs 3 and 16). A total of 1284 axes were counted, with 35 to 245 logs (mean of 128) in each accumulation. Ten morphotypes were identified based on anatomical characteristics and preservation state, and were assigned to lycopsids, sphenopsids and cordaitaleans with varying degrees of confidence. A fourth grouping, ‘probable cordaitaleans’, may include a small proportion of leaf petioles and stems derived from tree ferns and pteridosperms. However, the vast majority of these axes probably represent cordaitalean branches, because they lack traces of leaf petioles typical of tree ferns and the recurved branches that characterize pteridosperms. Plant fragments were assigned to one of five size fractions (<5 cm, 5 to 10 cm, 10 to 20 cm, 20 to 40 cm and >40 cm) based on their maximum preserved width. Although compaction and distortion has affected most stems, experimental analyses indicate that the width of a flattened axis yields a reasonable estimate of its original (pre-compaction) diameter (Walton, 1936; Niklas, 1978; Rex & Chaloner, 1983). The bulk of the material is <40 cm wide, with the 20 to 40 cm fraction having the most fragments (Fig. 16). In addition, the length of the longest log was recorded in 122 beds, with woody debris segregated into short (<20 cm), medium (20 to 50 cm) and long (>50 cm) size fractions (Fig. 17A). Axes 50 to 150 cm long are most common. The longest log observed was over 6 m, but historical records noted a specimen 18 m long from Grindstone Island (Emmons, 1836; Fig. 2).

Figure 16.

Size (width) distribution of 1284 fragments of woody debris segregated into four plant groups, based on measurements at 10 log accumulations (BP-1 to BP-10) in the Boss Point Formation (Fig. 3). Note that the grouping ‘probable cordaitaleans’ may contain rare tree fern and pteridosperm remains (see text for details).

Figure 17.

Maximum fragment size and woody debris abundance in the Boss Point Formation. (A) Size (length) distribution of longest fragment of woody debris recorded in 122 beds. (B) Proportion of woody debris versus sediment recorded in 88 beds. Woody debris accumulations most commonly constitute 5 to 10% of the bed but locally exceed 30%.

The quality and mode of preservation of woody material is variable. The smallest debris, from the bedding surfaces of mudstone and fine-grained sandstone units, consists of comminuted plant hash, coalified wood chips, twigs and bark fragments, and rare charcoal. Larger branches and trunks lie along or cross-cut bedding surfaces of medium-grained sandstone to gravel units and are decorticated and mainly flattened. Sediment-filled, coalified and/or partially silica or calcite-permineralized axes are oval in cross-section, and scattered, fully permineralized logs retain most of their original cylindrical shape. The original height of trees that contributed large woody debris to the fluvial system has been estimated using an equation developed by Niklas (1994) for the allometric relation between the diameter and height of extant trees:

display math(1)

where H is predicted fossil plant height and D is known stem diameter (all measurements in metres). Relevant only to woody plants, the equation was applied to the cordaitaleans, which were woody, and the lycopsids, which had little true wood but had thick, wood-like bark (periderm). Given that logs could have been derived from anywhere along the living tree trunk, the widths of flattened axes (or diameters of uncompacted specimens) provide rough minimum approximations of the original heights of cordaitalean and lycopsid trees in the log accumulations.

Plant groups and palaeoecology

Arborescent lycopsids, which had an ecological preference for the wettest available habitats, had a pole-like architecture for much of their life and had sparse lateral branches or developed a branching, reproductive crown just prior to death (Phillips & DiMichele, 1992; DiMichele & Phillips, 1994). The plants grew rapidly to attain heights of 40 m and were supported by a thick rind of decay-resistant, wood-like periderm. Lycopsid remains are very rare (2·5% of counts) in the succession and mainly comprise fragments of coalified bark, small branches and allochthonous roots (Stigmaria) 5 to 10 cm in diameter (Fig. 16), with most debris <60 cm long. The largest log observed was ca 30 cm in diameter and several metres long (incompletely exposed); a log of this diameter suggests a living tree ca 22 m high. Sediment-cast, in situ stigmarian rhizomorphs occur locally in floodplain deposits associated with thin carbonaceous shales. Several lycopsid genera probably contributed to the woody accumulations, but the only identifiable remains belong to taxa that preferred stable, peat-forming habitats (for example, Lepidodendron and Lepidophloios), rather than genera known to tolerate disturbed wetland habitats (for example, Sigillaria, Paralycopodites and Diaphorodendron) (cf. Calder et al., 2006).

The only sphenopsid remains encountered are Calamites, a ‘reed-like’ plant that, at least in clastic floodplain habitats, formed dense, clonal groves comprising jointed, hollow aerial shoots several metres in height (DiMichele & Falcon-Lang, 2012). The stems, which are uncommon (9·9% of counts), are <10 cm in diameter (Fig. 16) and reach 40 cm in length. Most axes are sediment-cast, and some are surrounded by a thin coalified rind. Transported stems are partly flattened, whereas rare upright examples remain cylindrical and typically are entombed in sandstone sheets in floodplain deposits. The type of Calamites observed was adapted to growth in shifting and rapidly aggrading sediment in disturbance-prone fluvial settings (Gastaldo, 1992; Pfefferkorn et al., 2001; Calder et al., 2006), particularly along channel margins, atop crevasse splays and in shallow standing water of abandoned channels and floodplain lakes (Scott, 1978; Bashforth et al., 2011).

As noted above, a small fraction of the remains assigned to ‘probable cordaitaleans’, particularly smooth to striate, flattened and coalified axes <10 cm in diameter (Fig. 16), may represent defoliated frond petioles and trunks of tree ferns and pteridosperms. Tree ferns, which were rare on Early Pennsylvanian landscapes (Pfefferkorn & Thomson, 1982), reached a few metres high, had a stem supported by a thick mantle of adventitious roots and were topped by a crown of pinnate fronds (DiMichele & Phillips, 2002). The opportunistic ferns rapidly colonized disturbed habitats or newly exposed substrates, and because they were free-sporing and required damp conditions to complete their reproductive cycle, were obligated to wetter settings. The pteridosperms encompassed a range of growth habits, but arborescent forms consisted of a robust, monoaxial stem supported by central wedges of wood and produced a crown of large, seed-bearing pinnate fronds (DiMichele et al., 2006). Most were slow-growing with a low reproductive output, and those in riparian habitats preferred nutrient-rich, clastic substrates in undisturbed settings, such as stable channel margins and distal interfluves (Bashforth et al., 2010, 2011).

Cordaitaleans were slow-growing and long-lived gymnosperms that were characterized by dense, conifer-like wood and a septate pith (Artisia). Forms that inhabited clastic substrates could achieve heights of 50 m, had a crown of large branches and produced extensive lateral root systems (Falcon-Lang & Bashforth, 2004, 2005; Falcon-Lang, 2006), which implies prolonged growth in stable settings where habitat disturbance was minimal (Bashforth et al., 2011). The gigantic cordaitaleans preferred moisture-deficient soils that developed in seasonally dry precipitation regimes (Falcon-Lang, 2004; Falcon-Lang et al., 2009; DiMichele et al., 2010; Bashforth et al., in press), occupying well-drained floodplains, channel margins and slopes bordering basinal areas (Falcon-Lang et al., 2004; Bashforth, 2005; Falcon-Lang, 2006; Libertín et al., 2009).

Cordaitalean remains are the principal vegetation type in the Boss Point Formation, making up 67·9 to 87·6% of all records and dominating all size categories (Fig. 16). Axes <10 cm in diameter mainly represent branches, whereas most logs >10 cm wide comprise trunks. The majority of axes are 10 to 40 cm wide, representing trees 10 to 25 m high, and the widest log measured in the 10 horizons analysed ranged from 45 to 90 cm in diameter, corresponding to living trees 27 to 37 m high. The largest fragment observed was 6·55 m long. Almost all axes are decorticated and have a smooth to ropey external surface and generally consist of a somewhat flattened cylinder of coalified wood that may be cored by a sediment-filled septate pith (Artisia) <5 cm in diameter. Extensive and well-preserved root systems are commonly found at the base of prone trunks. Small pods of convoluted sandstone sit within the coalified wood of some large logs and represent sediment-filled hollows (possibly due to wood rot) that became distorted during compaction and coalification. Partial permineralization by silica or calcite in some stems reveals the dense grain of the original Dadoxylon-type wood, and fully permineralized logs (up to 35 cm in diameter) first appear in megacycle 7 (Fig. 10), becoming more frequent higher in the formation. Although the majority of cordaitalean remains in the succession have been transported, in situ examples in fine-grained floodplain deposits include coalified stumps that retain part of the lateral root system (Fig. 15A), and networks of vertically oriented roots, either coalified or partially decomposed with drab reduction halos (Fig. 15C and D). Upright trunks are rare in sandstone bodies but include sediment-filled boles rimmed by a thin rind of coal (Fig. 15B), representing trees that grew atop channel bedforms or along channel margins and had rotted-out cores prior to burial. The fact that many cordaitalean trunks show evidence of wood rot, a characteristic of old trees, suggests derivation from mature forests.

Interpreted habitat preferences and effect on fluvial dynamics

The bulk of woody debris in modern fluvial channels is derived from riparian vegetation that collapses due to bank undercutting or erosion during major floods, lateral channel migration or avulsion (Scheihing & Pfefferkorn, 1984; Murphy & Koski, 1989; Spicer, 1989; Latterell & Naiman, 2007). Likewise, local riparian sources are inferred to have provided the bulk of the woody material in the Boss Point Formation.

It is believed that much of the braided-river plain was blanketed by dense stands of mature cordaitalean forest. This interpretation is supported by the large proportion of transported cordaitalean remains in channel sandstones and the presence of in situ cordaitaleans in both floodplain and channel deposits. The fact that intact root systems are preserved on many prone trunks further implies local derivation, because transport over long distances would have resulted in erosion of the thin lateral roots (cf. Falcon-Lang & Bashforth, 2004). The predominance of large cordaitaleans is consistent with deposition under a subhumid to semi-arid precipitation regime, the preferred climatic mode of these plants. Although the braided-river plain would have experienced extensive flooding and variable discharge during the wet season, the large size of the cordaitalean logs indicates that the forests persisted for decades to centuries and were largely unaffected by (annual?) floods. The trees probably dominated floodplain areas and were equally prominent in less-disturbed parts of active channel tracts, where they occupied stable channel margins and well-established vegetated islands. Considering its wide trunk, extensive branching crown and well-developed lateral root system, the incorporation of a large cordaitalean tree into a braided channel would have had an enormous impact on flow dynamics, resulting in substantial channel blockage, the build-up of sediment and woody debris, and possibly avulsion (cf. Gibling et al., 2010).

Calamites groves probably dominated the most disturbance-prone habitats, such as low-elevation bedforms and channel margins in active fluvial tracts, the shallow waters of seasonally flooded abandoned channels and crevasse splays on floodplains. Although the aerial shoots of Calamites were thin and only a few metres in height, dense thickets would have been capable of trapping and binding abundant sediment. As such, Calamites probably were fundamental in maintaining the cohesion of channel banks and, given that the group was predominant in early stages of vegetation succession (Calder et al., 2006; Bashforth et al., 2011), these ‘reed-like’ plants may have played a crucial role in the initial establishment and build-up of vegetated islands in active channels.

Although the braided-river plain accumulated under seasonal conditions, the presence of rare lycopsid remains, along with probable scattered tree ferns, implies that a variety of wetland habitats existed. In particular, the observation of transported Lepidodendron and Lepidophloios bark in channel sandstones and autochthonous stigmarian rhizomorphs in floodplain deposits indicates that ephemeral peat-forming mires dotted interfluve areas, even though coal seams are thin and exceedingly rare. The addition of distal floodplain elements into active channel systems suggests that channels occasionally avulsed through floodplain areas. A large lycopsid tree entrained in a narrow or shallow braided-river channel would have caused significant flow reduction and channel blockage, particularly a mature plant with a branching crown.

Types of Woody Accumulation

A new architectural element, WD, is described here in detail. This element consists of a heterogeneous interbedding of clastic sediment and abundant woody debris. Decomposition of organic material during burial brought about chemical reduction and precipitation of microcrystalline pyrite cement, and weathering of exposed surfaces produces a brownish-yellow colouration that makes these elements distinctive in outcrop (Fig. 7). The proportion of woody material within beds was estimated for 88 bedsets using the visual comparative tables of Terry & Chillingar (1955). Values range from 1 to 40%, averaging 13·1 ± 9·5%. Most occurrences constitute 5 to 10% of the bedset (Fig. 17B) and, as fragment size increases, the proportion broadly diminishes.

Although a few logs are filled with sediment, the great majority of woody fragments are flattened. Al-Silwadi (2011) estimated the degree of compaction experienced by 1·2 m of wood-bearing sandstone at the 187 m level (Fig. 6) using the method of Gibling et al. (2010). The method assumes that the width of flattened fragments approximates their original diameter, as noted above, and that sandstone compacts only slightly, with a decompaction ratio of 1·1. The width of 150 fragments ranged from 0·3 to 16·2 cm, with 60% of the fragments >1 cm wide, and the coalified thickness of a set of fragments averaged 2·9 mm. Dividing the aggregate width of the fragments (original diameter) by their aggregate coalified thickness, a decompaction ratio of 7·1 times was obtained. Using software analysis of three photographs of bed cross-sections, the proportions of woody debris and sandstone were estimated at 21·3% and 78·7%, respectively. Applying these proportions to decompaction ratios of 7·1 for the wood and 1·1 for the sandstone, the bed would originally have been 2·85 m thick, some 2·4 times thicker than at present.

Accumulations of woody debris are widespread in the formation, generally concentrated at the bottom of braided-fluvial units and particularly prominent in megacycle 3 (Fig. 18). Three types of WD were recognized: lags (WD-a; 17 occurrences, 28% of all the WD recognized); channel fills (WD-b; 23 occurrences, 38%); and macroform cores (WD-c; 21 occurrences, 34%). A few associations of WD with evidence of in situ plant colonization may represent components of vegetated islands, but the scattered nature of these occurrences prevented systematic counting.

Figure 18.

Stratigraphy and tidal platform tracing of part of the braided-fluvial unit of megacycle 3 on the tidal platform (enlarged from Fig. 10), showing archetypal facies successions and architecture for the three types of woody accumulation (reported in the white outlined squares and with major woody debris reported in the tidal platform tracing). Lag accumulations interbedded with gravel bars and bedforms are reported in (A); a channel fill accumulation occupying the lower portion of a minor channel is reported in (B); a macroform core accumulation interbedded with laminated sand sheets is reported in (C). For abbreviations, see Table 2; for symbols, see legend in Fig. 4.

Because terrigenous sediment is the predominant component, the WD occurrences could be assigned to previously described elements with similar internal architecture (for example, GB, CH and DA). However, as discussed below, the prominence of wood points to distinctive depositional processes and settings, justifying the recognition of WD as a new element. Moreover, the significance of woody debris in the morphodynamic evolution of ancient rivers has been undervalued and deserves emphasis, especially where abundance and preservation of woody debris is exceptional.

Lags (WD-a)

Observations

The lags form sheets that rest on prominent erosional surfaces, commonly the basal surfaces of channel bodies (285 to 293 m in Fig. 18A; Fig. 19A and B). The lags range from 15 to 50 cm in thickness and exceptionally thick examples may reach 2 m; they usually lie on sixth-order surfaces, are topped by third-order surfaces and include second-order surfaces. The woody material is interbedded with clast-supported massive gravel and trough cross-bedded sandstone, rarely with planar cross-bedded gravel, and the bedding has been deformed by differential compaction (cf. Gibling et al., 2010).

Figure 19.

Woody accumulations of the WD elements. (A) Lag accumulation (WD-a) ca 2 m thick, resting on channel base (arrowed) and extending up to dashed line. (B) Close-up of logs in the accumulation shown in (A). (C) and (D) Channel-fill accumulations within minor channels (WD-b) at the top of thick channel deposits, with base of channels arrowed. In (C), the fill is ca 4 m thick with logs throughout, extending up to the dashed line and overlain by shale of an abandoned-channel fill. In (D), the fill is up to 3 m thick, cut into thick channel sandstone with base at dashed line and is overlain by shale. (E) and (F) Macroform core accumulations (WD-c). In (E) [close-up in (F)], a narrow accumulation of logs 1·5 m thick is flanked by sandstone with centroclinal cross-strata directed towards the accumulation. The logs form the core of a mounded macroform 2·5 m thick with cross-strata directed away from the core at higher levels. Above, the fill of a secondary channel rich in logs (WD-b) has a planar cross-set and plane beds in its topmost part below abandoned-channel shale. (G) and (H) A large sediment-filled log [close-up in (H)] forms part of a woody accumulation 50 cm thick, overlain by mounded, cross-bedded sandstone.

Woody debris consists of logs, bark and finely comminuted fragments and, based on a detailed analysis of two lags (Fig. 20A), most phytoclasts are <10 cm wide, although rare examples reach 60 cm wide and 6 m long. Cordaitaleans predominate (56 to 81%) but sphenopsids and lycopsids together comprise 19% of 323 counted specimens (Fig. 20A). The woody material lacks preferential alignment. Lag thickness does not appear to scale with the channel size, based on comparison between the size of woody debris and the thickness of fining-upward cycles. The deformation generated by differential compaction has locally resulted in the concentration of woody debris as centimetre-thick horizons with a lateral extent of ca 10 m.

Figure 20.

Size (width) distribution of woody fragments segregated into four plant groups, based on measurements in two lag (WD-a) (A), three channel fill (WD-b) (B) and five macroform core (WD-c) (C) accumulations.

Interpretation

The accumulations mainly rest on the bases of major channels, represented by sixth-order surfaces, where prolonged erosion would have resulted from confined turbulent flows (Dietrich et al., 1989). The lags are inferred to have been deposited following episodes of erosion and undercutting of channel banks. The relative abundance of lycopsids and sphenopsids suggests that undercutting, probably triggered in some cases by avulsion (Gibling et al., 2010), intersected poorly drained floodplains and disturbance-prone habitats. The common association of wood with gravel indicates that the woody fragments were part of channel lags (Parker & Sutherland, 1990), probably deposited and subjected to armouring soon after the undercutting phase (Gomez, 1983). The lack of preferred alignment of the wood fragments may reflect rapid deceleration that favoured quasi-instantaneous obstruction, as well as interference between fragments that were heterogeneous in size and shape (Hubert, 1967), consistent with the minimal size sorting of the debris. The presence of large branching crowns or root systems on entrained trees may have enhanced these processes. The fragments probably continued to accumulate on the channel base until they were buried with sediment, and the third-order surfaces above some lags suggest that they contributed to channel-base macroforms created by dune trains.

Channel fills (WD-b)

Observations

Channel-fill accumulations are lenses resting on concave-up or irregular erosion surfaces that commonly define minor channels (Fig. 7; 293 m and 298 m in Fig. 18B). The channel fills typically lie at the top of thick channel-sandstone bodies, into which they are incised, and they are overlain by shale (Fig. 19C and D). Accumulations range from 50 cm to 2 m in thickness and rest on fifth-order surfaces. Locally, larger channels represented by sixth-order surfaces are completely filled by up to 6 m of wood-rich sediment (Fig. 7). The accumulations terminate at fourth-order or higher-order surfaces, and they include third-order surfaces. The woody debris is interbedded with irregular sheets and lenses of sandstone (Fig. 19C) or lies within fining-upward rhythms of trough cross-bedded gravel and sandstone that locally pass upward into plane-laminated and ripple cross-laminated sandstone. Convolute bedding is more frequent near the base of the accumulations.

Woody debris consists of logs, bark and finely comminuted material. Detailed analysis of three accumulations indicates that 10 to 40 cm wide phytoclasts are most common (Fig. 20B), the overwhelming majority of which were derived from cordaitalean trees (80 to 92%), with rare lycopsid (2%) and sphenopsid (6%) axes. Fragment size typically decreases upward within accumulations, following the grain-size trend of the clastic host. No preferential orientation was observed.

Interpretation

Accumulations are interpreted as the fills of abandoned or partially obstructed minor and, more rarely, major channels; their preferential position at the top of thick channel-sandstone bodies suggests that the channels were small conduits at high topographic levels within major channel belts. The prevalence of cordaitaleans indicates that, during active phases, channels mainly cut through the stable margins of dense, mature forest. The occurrence of larger fragments suggests short transport and limited physical degradation of the debris, consistent with the preservation of root stocks at the base of many cordaitalean logs. Fining-upward rhythms indicate progressive waning of flow, and the shift from dunes to plane beds and rippled beds suggests shallowing of flow during channel filling (Stear, 1983). The association of larger fragments with gravel implies that accumulation of woody debris commenced during an active stage of the channel, with the main phase of accumulation perhaps triggered by a high-magnitude flood or by the sinking of saturated and degraded debris (Gastaldo & Degges, 2007). The lack of preferred orientation may reflect interference between fragments with varied size and shape. Deposition of highly degraded fragments after extended transport is excluded by the common occurrence of root stocks at the base of cordaitalean logs. The build-up of woody material in smaller channels would have reduced the channel capacity and induced sediment accumulation (Curran & Wohl, 2003), which may have promoted overflow, avulsion and channel abandonment (cf. Gibling et al., 2010). Fourth-order surfaces above many accumulations imply that their topmost parts formed part of macroforms within the channels, and higher order surfaces above indicate that the channel fills were truncated by later channel-forming events.

Macroform cores (WD-c)

Observations

Woody accumulations form prominent mounds in the cores of some planar or gently mounded macroforms (Fig. 8; 325 to 330 m in Fig. 18C), with subordinate plant material at the macroform top. The accumulations rest on third-order surfaces and are topped by, or located just below, fourth-order surfaces. Accumulations include internal third-order surfaces and are chiefly hosted in trough cross-beds that pass up into planar cross-bedded and plane-laminated beds, with minor ripple cross-laminated sets at the top of a few accumulations. Accumulations are well-stratified with minor, localized convolute bedding and occur at the top of major sandstone bodies.

In a particularly informative example (Fig. 19E and F), a narrow body of stacked woody material and sandstone forms the core of a sandstone mound 2·5 m high and at least 10 m wide, interpreted as a macroform. Cross-strata within an adjacent scour dip towards the core (centroclinal cross-stratification), whereas overlying beds dip in the opposite direction (Fig. 19E). A channel body cuts the mound with its lowest point above the core, suggesting that the plant remains compacted during shallow burial. The channel, an example of element WD-b, contains woody debris and sandstone, and is overlain by a planar cross-set capped with plane beds below shale. In another example (Figs 19G and H), a sheet of woody material and sandstone contains a log 75 cm wide, partially filled with sediment, with a mound of cross-bedded sandstone and shale (macroform) centred over the log.

Detailed analysis of five accumulations indicates that woody fragments encompass a wide range of sizes, with most <40 cm wide (Fig. 20C). Lycopsids compose 2·5% of the debris and sphenopsids reach 10%, but the bulk of the material is cordaitalean (64 to 87%). The larger logs are oriented mostly parallel and, less commonly, perpendicular to flow, forming the core of mounded bodies. The flow-parallel orientation is particularly evident in examples illustrated in Figs 8 and 9. In contrast, medium-sized debris is chaotic in attitude. These accumulations range in thickness from 1 to 4 m.

Interpretation

The close association of woody material with sediment macroforms bounded by third-order and fourth-order surfaces suggests that sediment accreted around debris on the channel floor and subsequently extended upstream and downstream (Fig. 19G). The upward transition from trough cross-beds to plane-laminated stratification suggests that upper regime flows developed during shallowing (Stear, 1983). Associated ripple cross-laminated sets may represent decelerating flow or conditions close to the ripple-plane bed-antidune stability triple point. Some accumulations are capped by sandstone, indicating burial by other downstream-migrating bars. Accumulations of sand and phytoclasts near the tops of channel fills would have contributed to progressive shallowing and eventual macroform deactivation due to flow diversion (Gibling et al., 2010).

The macroform core accumulations probably originated after standing trees toppled into the channel or after large transported logs, such as the example illustrated in Fig. 19G, were stranded, with subsequent trapping of additional floating debris (Abbe & Montgomery, 1996). This process was probably enhanced by the crown morphology of cordaitaleans and occasional lycopsids that entered the river. The relatively high proportion of sphenopsids suggests that some core debris originated from disturbance-prone sites, possibly channel margins or vegetated islands. The preferential flow-parallel orientation suggests tractional reworking, and logs oriented perpendicular to flow probably migrated by rolling and were arrested in low-velocity zones. MacDonald & Jefferson (1985) reported comparable trends in flume experiments, although clustering of debris may also favour preferential orientation (Blair, 1987).

Evidence for Vegetated Islands

Vegetated islands are a common geomorphic component of many modern braided rivers (Desloges & Church, 1987; Gurnell et al., 2002, 2002; Francis et al., 2009). However, their recognition in the stratigraphic record of modern or Holocene fluvial deposits has been difficult, due to scale factors or difficulties in performing detailed geophysical surveys (Horn et al., 2012a). To the knowledge of the present authors, vegetated islands have not previously been recognized in ancient fluvial deposits. In modern braided rivers, vegetated islands form through stabilization of accretionary macroforms, or through dissection and isolation of floodplains near the active channel belt (Kollmann et al., 1999). The stratigraphic record of stabilized macroforms and dissected floodplains is radically different (cf. Francis et al., 2009), and the exceptional Boss Point exposures allow the mechanism of vegetated-island formation to be investigated.

The process of accretion of a vegetated island within a channel belt comprises several stages (Kollmann et al., 1999): (i) accumulation of barren macroforms and woody debris; (ii) colonization by opportunistic, pioneering plant types; and (iii) final stabilization by slow-growing, site-occupying vegetation. This pattern exemplifies a continuous succession of intermediate vegetation stages, as the ecosystem proceeds towards a ‘steady state’. The first phase of accretion implies a progressive shallowing in the areas of incipient island formation, associated with a gradual flow diversion towards deeper portions of the channel belt. The subsequent phases of plant colonization record increasing occurrences of standing vegetation with deep and laterally extensive root systems. Some of these features are observable in the braided-fluvial units of the Boss Point Formation, implying the presence of vegetated islands. A few associations of WD elements, usually represented by composite lags and macroform cores, are associated with diffuse rooting, exceptionally (five occurrences) with standing vegetation. Other lines of evidence, such as vegetation-induced sedimentary structures (Rygel et al., 2004) and plant debris wrapped around tree stumps or laid-over trunks still attached to their root systems (Fielding et al., 2011), are lacking, probably due to their limited preservation potential.

Figure 21 illustrates the lowermost portion of megacycle 3, which provides the clearest evidence for a vegetated island in the Boss Point section. Above the base of the braided-fluvial unit (250 m, Fig. 21), sandy and gravelly bedforms contain lags with abundant woody debris (WD-a) and an upright cordaitalean tree but no evidence of roots (Fig. 15B). From 253 to 258 m, trough cross-bedded sandstone with woody debris passes upward into plane beds and antidunes with common roots and an upright cordaitalean tree. This interval, which is 8 m thick, is inferred to reflect a transition from bedforms within a channel to a vegetated island, on the basis of its fining-upward trend, the appearance of plane beds, and the presence of roots and standing trees. The rooted interval is cut by a minor channel (fifth-order surface) containing bedforms and woody debris (WD-b), the latter possibly derived from erosion of the vegetated island. This candidate vegetated island is continuous in the tidal platform for up to 70 m in a direction parallel to the mean flow (Fig. 21). A sixth-order erosional surface tops the succession and eventually cuts through it with an erosional relief of ca 10 m, suggesting that, in its late stages, the vegetated island acted as a stabilized terrace bordering a major active channel.

Figure 21.

Stratigraphy and tidal platform tracing of the basal portion of the braided-fluvial unit of megacycle 3, showing the architecture of a putative vegetated island (enlarged from Fig. 10). Standing trees and rooted horizons are reported for clarity also on the shore tracing. Shallowing-upward successions are composed of interbedded WD and woody debris-free elements, with rooting and standing vegetation situated above woody accumulations. The vegetated island is cut by a major, sixth-order erosional surface that probably represents the margin of a major channel. For symbols, see legend in Fig. 4.

Discussion

The Boss Point Formation provides a rare opportunity to study the interactions between vegetation and fluvial dynamics in the fossil record. Woody accumulations represent a significant portion of the formation and are inferred to have had a considerable effect on depositional processes, as in modern island-braided rivers with abundant vegetation (Gurnell et al., 2000; Coulthard, 2005; Collins et al., 2012). In addition, the paucity of pristine modern fluvial systems makes the identification of fossil analogues difficult (Gastaldo & Degges, 2007) because vegetation has rarely been linked to the architecture and stratigraphic signature of modern rivers (with a few exceptions such as Horn et al., 2012a).

Role of vegetation in the Boss Point depositional system

The abundance and diversity of woody debris implies a densely vegetated landscape and a sustained supply of large plant fragments to the fluvial system. The three WD subtypes represent different mechanisms of phytoclast accumulation on a small to moderate scale, as sheets on the floors of major channels (WD-a), fills of minor channels (WD-b) and macroform cores (WD-c). On a larger scale, interactions with in situ plants probably also promoted the accretion of stable vegetated islands.

Although the abundance of fossil woody material implies that the Boss Point plain was densely vegetated and that channel banks were frequently undercut, inferred vegetated-island deposits are rare in the formation, suggesting that migration, avulsion and reoccupation of channel belts largely destroyed them. Deposits associated with the inferred islands comprise compound accretionary macroforms laid down in channels and subsequently penetrated by roots, but the inferred island deposits lack overlying floodplain fines, which were not deposited, or have been eroded. The association of standing trees (Fig. 15B) with some WD elements within channel deposits suggests that the trees acted as local dams (Alexander et al., 1999; Nakayama et al., 2002) and assisted in the establishment of pioneer vegetation (Rodrigues et al., 2006). Sediment capture and build-up that initiated island formation probably started high in the channels due to colonization by dense groves of opportunistic and disturbance-adapted calamitaleans, which are common in the cores of macroforms (element WD-c). As sediment aggraded and vegetated islands became increasingly elevated and less flood-prone, slow-growing and long-lived cordaitalean trees (Fig. 15B) would have gradually taken over as the dominant vegetation type in ‘stable state’ communities.

The abundance of woody debris in the fills of minor channels suggests that these smaller conduits were prone to being choked by plant remains transported during floods, as in some modern rivers (Makaske et al., 2002). In addition, banks may have been better protected by vegetation in smaller channels, aiding channel filling: Church (2002) and Eaton & Giles (2009) noted that bank strength due to vegetation influences the geometry of small channels but has little effect on larger channels.

A significant portion of the Boss Point phytoclasts probably entered the system from the erosion of vegetated islands, channel margins and adjacent or upstream floodplains, from which large quantities of woody debris are introduced to modern rivers through floods, bank erosion and windthrow of trees (Palik et al., 1998). A source of uncertainty is represented by the difficulty in estimating the transport distance of woody debris. The association of logs with reworked palaeosol materials or remnants of collapsed floodplain blocks has been linked with short transport distance (Liu & Gastaldo, 1992), whereas an increase in physical degradation of woody debris suggests longer transport. In the Boss Point Formation, woody debris in lag accumulations within large channel bodies (WD-a) is associated with reworked palaeosol nodules, suggesting a limited transport distance and early bed armouring (Gomez, 1983). The accumulations within minor channels (WD-b) also suggest limited transport of phytoclasts in minor conduits (Marcus et al., 2002). Furthermore, the preservation of root stocks at the base of many cordaitalean logs indicates limited transport. Thus, much of the vegetation in the Boss Point rivers appears to have been locally derived, affecting the depositional style of the formation and, in turn, influencing the facies model.

Facies model

The Boss Point Formation can be subdivided into distinct facies associations, representing the aggradation of floodplains and lakes; the migration of macroforms and formation of vegetated islands with subsequent episodes of accretion and erosion; and the scouring, filling and avulsion of braided channels (Fig. 22).

Figure 22.

Architectural hierarchy and facies model of the Boss Point Formation, with the four main archetypal facies successions recognized. Sandstone-dominated units represent braided rivers, the evolution of which is governed by processes of bar and island accretion and migration (A) and channel organization (B). Mudstone-dominated units represent vegetated floodplains subject to pedogenesis (C) and with shallow anoxic lakes with mixed clastic, carbonate and organic deposition (D). Abbreviations for architectural elements are as follows: gravel bars and bedforms (BG), sandy bars and bedforms (SB), laminated sand sheets (LS), downstream-accretionary (DA) and lateral-accretionary (LA) macroforms, rippled stacks (RS), overbank fines (OF) and crevasse splays (CS).

The braided tracts show a complex architecture that contrasts with the tabular floodplain deposits. The deposits of vegetated islands (Fig. 22A) alternate with those of active and abandoned-channel fills (Fig. 22B). Stacked successions of trough cross-sets and minor planar cross-sets (SB) are the predominant component of channel fills and comprise 55% of the in-channel deposits; they represent fields of two-dimensional and three-dimensional dunes of varied scale (Browne & Plint, 1994). Where represented by planar cross-sets and bedform successions linked to vegetated islands, sandy bars and bedforms are commonly associated with macroform core woody accumulations (WD-c) and capped by sand sheets that represent progressive shallowing high in the channels (Stear, 1983). Within major and minor channel fills, the sandy bars and bedforms are associated with gravel dunes and barforms (GB), which constitute less than 5% of the in-channel deposits, and woody lag accumulations (WD-a), formed low in the channels. Minor channels are filled by woody debris and rippled sandstone (WD-b and RS), which represent late stages of channel filling and abandonment. Occurrences of RS constitute 18% of the in-channel deposits and are preferentially associated with antidunal bedding (LS) towards the top of the formation. Successions of ripple-bedded and antidunal-bedded strata probably represent sharp discharge variations induced by enhanced seasonality (Allen et al., 2011). Notably, downstream and lateral-accretion macroforms (DA and LA), the product of larger macroforms, are strongly subordinate, constituting in combination with LS less than 10% of the in-channel deposits, in contrast with the 18% of accumulations of woody debris. Prior to compaction of the wood, the proportion of woody accumulations would have been considerably greater, as indicated by the decompaction assessment, although some compaction may have taken place shortly after burial (see discussion of Fig. 19F).

Analysis of a small number of modern sandy braided streams over varied durations suggests that their planform configuration is chiefly a function of discharge cycles on time scales of a few years, commonly a decade or less (Jones, 1977; Cant & Walker, 1978; Blodgett & Stanley, 1980; Horn et al., 2012a). In these cycles, high-flood stages are associated with undercutting and landscape instability, whereas bar migration and channel filling take place during the waning stages, when large macroforms are progressively reworked into smaller bars (Horn et al., 2012b). Gradual exposure of channel floors and extensive stabilization to form vegetated islands takes place during low-flood stages (Joeckel & Henebry, 2008), and the vegetated tracts are mostly degraded rapidly during the successive high-flood stage. The smaller unit bars and channel fills have the highest long-term preservation potential and may represent the predominant stratigraphic signature of the fluvial system.

Well-drained floodplains are dominated by sheets of oxidised fine-grained deposits (Fig. 22C), and subordinate sandstones represent small, isolated channel fills and rare splay systems. The characteristics of the channel fills are typical of anabranching high-accommodation systems (Davies & Gibling, 2003; Rygel & Gibling, 2006). Splay systems are rare in the lowermost 1000 m but examples appear in the upper 200 m. The paucity of splay systems, at least in the Boss Point type area, could be explained in several ways: a densely vegetated environment may have impeded floodwaters and trapped sediment (Tooth & Nanson, 2000); bordering levées may have limited overspill, as represented in the model of deep, perennial, sand-bed braided rivers of Miall (1996); or regional avulsive events may have reduced the supply of coarse sediment to floodplains (Cain & Mountney, 2009). The appearance of splay systems in the uppermost strata may reflect a shift towards more arid conditions with intensified seasonal flow events (Calder et al., 2005; Allen et al., 2011). Lacustrine and poorly drained floodplain tracts are represented by stacked, drab fine-grained sheets of sandstone and shale with coal, carbonaceous shale and limestone (Fig. 22D). Browne & Plint (1994) inferred the existence of extensive and rather long-lived lakes, based on the lateral traceability of these horizons.

The results herein accord with those of Horn et al. (2012a) for the Platte River of Nebraska (USA), a river that has been considered as a reference for sandy braided-fluvial facies models (Horn et al., 2012a). The overwhelming majority of the 1200 m thick Boss Point Formation consists of medium-size sandy bedforms and bars (largely trough cross-beds), and large downstream and laterally accreted elements are uncommon (Table 2). Furthermore, where the architecture can be discerned, discrete channel fills that represent minor channels are common, suggesting that the filling and abandonment of small anabranches was an important process and that the resulting sediments were readily preserved.

The prominence of both major and minor channels, locally containing abundant woody debris, and the presence of successions attributed to vegetated islands suggests that the Boss Point rivers were anabranching and similar to some modern island-braided and wandering systems (Desloges & Church, 1987, 1989; Gurnell et al., 2000, 2001; Tockner et al., 2003; Davies & Gibling, 2013), or more broadly to sandy anabranching rivers (Jansen & Nanson, 2010). In sandy anabranching rivers associated with seasonal settings, in-channel vegetation is abundant and is commonly associated with downstream shadow bars. In the Boss Point Formation, however, standing trees were rarely preserved and no shadow bars were identified, hampering any direct comparison with modern dryland settings.

Comparison with modern rivers

In recent years, the integration of geomorphic and GPR investigations has allowed the linkage of planform configurations to the subsurface architecture of modern rivers (Lunt et al., 2004; Sambrook Smith et al., 2006), thus improving the comparison with ancient analogues. Discharge exerts a primary influence on river morphology (Jones, 1977), but reliable discharge estimates from examples in the rock record are often unattainable (Davidson & North, 2009). Moreover, human impacts have significantly altered fluvial dynamics in the last two centuries (Gastaldo & Degges, 2007; Walter & Merritts, 2008; Wohl, 2013). Despite these sources of uncertainty, architectural evaluations remain virtually the only method for comparing modern and ancient rivers.

The woody debris accumulations described in this study are comparable to those reported in the modern Yazoo Basin rivers (northern Mississippi, USA), which are unstable sand-bed floodplain streams adjacent to the Mississippi River. Here, woody fragments within the channels cause damming and channel obstruction (Wallerstein & Thorne, 2004), comparable to the channel fill and macroform core accumulations of the Boss Point Formation. This comparison supports the present observations that the Boss Point fluvial system was characterized by a wide variation in the ratio of woody debris size to channel width (Wallerstein et al., 1997). The presence in the Boss Point Formation of vegetated islands cored by woody accumulations suggests that barhead log jams could have been stabilized by riparian vegetation, as observed in the modern Tagliamento River of Italy (Coulthard, 2005) and the Burdekin River of Australia (Fielding et al., 2011). In the latter rivers, vegetated islands are characterized by dynamic and unstable planforms but exert primary controls on patterns of braiding in the channel belt, promoting the formation of anabranches of varied size (Kollmann et al., 1999; Francis et al., 2009).

The Boss Point braided alluvium is characterized by the abundance of medium-sized bedforms and woody debris accumulations, and by the scarcity of large macroforms. Channels probably ranged from 5 to 13 m in depth and 40 to 400 m in width. Modern analogues previously proposed for the Boss Point rivers, in terms of dimensions and architecture, include the Brahmaputra, South Saskatchewan and Platte rivers (Browne & Plint, 1994). However, these comparisons do not take into account the influence of vegetation and human impacts on the system, topics that require critical discussion (Bruijnzeel & Bremmer, 1989). For example, the abundance of medium-sized dunes in the Platte River is probably linked to channel shrinkage and stabilization by vegetation (Joeckel & Henebry, 2008). Woody debris in the Platte River was briefly reported by Tate & Heiny (1995) but detailed accounts are currently lacking, and GPR profiles, although informative across exposed sediment, could not be obtained across vegetated islands (Horn et al., 2012a). Similarly, large accumulations of woody debris are present in the South Saskatchewan River, but neither vegetation nor human impacts are discussed in existing facies models or architectural studies (Cant & Walker, 1978; Sambrook Smith et al., 2006; Parker et al., 2013). No comprehensive model currently exists for fluvial architecture in association with the influence of vegetation in modern pristine rivers, with the sole exception of Fielding et al. (2011), which highlighted the presence of upright trees within channels in a facies model for seasonal tropical rivers.

To the knowledge of the present authors, the only study that integrated woody debris accumulations with the planform configuration and GPR architecture of a modern sandy braided river is that of Skelly et al. (2003), which documented a shallow braided belt from 90 to 330 m wide, characterized by trough cross-bedding and subordinate planar cross-bedding in the Niobrara River (Nebraska, USA). Woody debris accumulations in channel deposits are represented by parabolic, convex-up radar facies that resemble the macroform core accumulations described here. Although a proper architectural comparison is biased by the different scale and abundance of woody debris in the Boss Point and Niobrara systems, Skelly et al. (2003) demonstrated how future studies have the potential to document the influence of vegetation on the stratigraphic signature of a modern river.

The influence of woody debris on a fluvial system also depends on the physical properties of the in situ parent vegetation. These aspects varied greatly through time, and a comparison between modern and ancient vegetation is recommended when dealing with examples in the rock record. This comparison is essential in Palaeozoic fluvial systems, the architecture of which is closely linked with the evolution of plants (Gibling & Davies, 2012).

Preservation of woody debris, and Palaeozoic plant evolution

The Boss Point phytoclasts underwent both coalification and permineralization. Coalification is an effect of pressure and temperature overprinting on the original tissue (Spackman & Barghoorn, 1966), with flattening of axes indicating strong compaction. Permineralization involves tissue replacement by calcite and silica, from saturated fluids in permeable deposits (Leo & Barghoorn, 1976). Permineralized logs preserve most of their original shape, indicating that this process took place soon after burial (De Lafontaine et al., 2011).

The abundance of woody debris in channels of the Boss Point Formation indicates favourable preservation conditions. Within the channels, rapid burial and onset of reducing conditions below the sediment surface would have hindered organic degradation. Preservation would also have been facilitated by a fairly high groundwater table under relatively humid conditions during much of the history of the formation (Allen et al., 2011), an inference supported by the presence of lake deposits and coals. Moreover, the depositional style probably favoured rapid channel diversion and the formation of short-lived islands (cf. Horn et al., 2012b), providing many opportunities for rapid burial of woody material.

The presence of free calcite and silica versus siderite governs early diagenetic processes, favouring permineralization versus coalification, respectively (Taylor et al., 1989; Williams et al., 2010). A shift towards semi-arid conditions influences the chemistry of shallow groundwater and results in siderite depletion (Dutta & Suttner, 1986). Accordingly, the upward increase in permineralized logs in the Boss Point Formation may indicate greater seasonality and a more arid climate (Allen et al., 2011), or a shift to a better-drained, more inland setting (cf. Gibling et al., 2010).

Although the first unequivocal evidence for higher plants dates to the Early to Middle Ordovician (Rubinstein et al., 2010), large vegetation appeared only near the Silurian–Devonian boundary where the oldest root structures are attributed to Prototaxites (Hillier et al., 2008). These early saprophytes (Hueber, 2001; Boyce et al., 2007) probably did not aggregate in densely vegetated clusters. In the Middle Devonian, patches of fern-like trees several metres high reached densities of 600 to 800 trees/hectare (Goldring, 1927; Moore, 1933; Driese et al., 1997), and woody material is common in upper Devonian strata. However, large woody material becomes abundant within fluvial channels near the Mississippian–Pennsylvanian boundary (Falcon-Lang & Scott, 2000) and is a prominent feature of Lower to Middle Pennsylvanian fluvial deposits (Gradziński et al., 1982; Falcon-Lang & Bashforth, 2004, 2005; Bashforth, 2005; Gastaldo & Degges, 2007; Gibling et al., 2010).

The large size and abundance of cordaitalean logs in woody accumulations in the Lower Pennsylvanian Boss Point Formation underscores the importance of this plant group in the evolution of fluvial systems, and suggests that, by the earliest Pennsylvanian, vegetation was abundant enough to strongly influence channel morphology. The appearance within channel belts of vegetated islands initially stabilized by pioneering, disturbance-prone taxa is a further key point of evidence for the ongoing co-evolution of fluvial styles and vegetation during the Carboniferous (Davies & Gibling, 2013).

Conclusions

In the Boss Point Formation of Atlantic Canada, abundant woody debris in channel deposits and roots in floodplain deposits testify to a densely vegetated landscape. Braided-fluvial deposits were dominated by the aggradation and lateral wandering of sandy braided channels, locally more than 10 m deep and up to 400 m wide, that occupied regional valleys up to 100 m deep and at least 12 km wide. The channel deposits contain sandy and minor gravel bars and bedforms. The preservation of large downstream-accreted and laterally accreted macroforms is relatively uncommon, but high flow-strength sand sheets and rippled abandonment intervals are locally prominent.

A new architectural element, woody debris accumulations (WD), comprises lags, channel fills and macroform cores composed of clastic sediment and abundant cordaitalean logs, with minor lycopsids and sphenopsids. These plant groups represent a landscape characterized by extensive mature interfluve forests, dotted by patches of disturbance-prone pioneer vegetation and subordinate peat-forming mires. Scattered upright plant remains in channel deposits suggest that stable vegetated islands were situated within braided channels. The preservation of exceptionally abundant woody material, a reflection of rapid burial below the water table in a relatively humid climatic setting, testifies to the crucial role that the large and architecturally complex cordaitalean trees played in the geomorphic evolution of Early Pennsylvanian landscapes. The woody accumulations are associated with particular orders of stratigraphic surface, with lags and channel fills resting on erosional channel bases (fifth-order and sixth-order) and macroform cores typically associated with lower order surfaces near the tops of braided-fluvial bodies.

The Boss Point Formation yields a remarkable Late Palaeozoic record of fluvial dynamics linked to vegetation. Deciphering these interactions is important for establishing realistic facies models for ancient fluvial deposits. The abundance and importance of vegetation in the formation raises questions about the validity of modern analogues based on landscapes extensively modified by human activity. Although the architectural analysis of modern rivers in vegetated landscapes needs more attention, the profound influence of vegetation on modern rivers has been documented, in particular, for the Yazoo Basin (Mississippi, USA), the Tagliamento (Italy), the Platte and Niobara (Nebraska, USA), and rivers of the Pacific Northwest and Australia. In this respect, the future integration of stratigraphic, geomorphological and geophysical approaches has considerable potential for understanding more deeply the influence of vegetation on rivers.

Acknowledgements

Rhea Hurnik and Brian Hebert offered crucial field support to this study. The Joggins Fossil Centre and particularly Jenna Boon are also acknowledged for helpful cooperation. The authors are greatly indebted to reviewers John Holbrook and Guy Plint for constructive comments that significantly improved the manuscript. Likewise, Associate Editor Nigel Mountney and Editor Stephen Rice are warmly thanked for their assistance. The study was supported by an International Association of Sedimentologists Post-Graduate grant to Alessandro Ielpi, a Discovery Grant from the Natural Sciences and Engineering Research Council of Canada (NSERC) to Martin R. Gibling, a Postdoctoral Fellowship from NSERC to Arden R. Bashforth, and an American Chemical Society Petroleum Research Fund Grant (#47967-GB8) to Michael C. Rygel.

Ancillary