1 Chemostratigraphic methods include those that characterize sedimentary successions based on distinctive features in their inorganic whole-rock geochemistry. Their applications have most commonly been in the subdivision of deformed red beds, where biostratigraphic and magnetostratigraphic methods are limited by lack of preserved microfossils and uncertain reorientation of magnetic minerals in the strata (e.g., Preston et al. 1998; Pearce et al. 1999, 2005, 2008; Ratcliffe et al. 2004, 2010).

2 Red beds are common in Upper Devonian to Pennsylvanian strata of southeastern New Brunswick, where they form part of the fill in a system of basins (Cocagne Graben, Moncton, Sackville, and Cumberland basins) that are components of the regional Maritimes Basin Complex (Fig. 1; Keighley 2008). Three tectono-sedimentary megacycles (“allocycles” of St. Peter 1993) have been identified (Fig. 2) that indicate a complex history of fault-controlled subsidence, strike-slip motion, basin inversion, and salt tectonics (e.g., Wilson 2005; Wilson and White 2006; Park and St. Peter 2009; Park et al. 2007, 2010; Waldron et al. 2013; Craggs et al. 2017). Red beds are particularly abundant in the 2^nd megacycle (Upper Mississippian), where they have been formally differentiated into several formations in the Moncton and adjacent basins. However, in the western part of the Moncton Basin, limited outcrop and borehole data to date have resulted only in a few broadly based studies of the Upper Mississippian red beds (Anderle et al. 1979; Wilson 2005), an informal lithostratigraphy at the sub-group level (St. Peter 1993), and continued uncertainty as to the presence or importance of a mid-red bed unconformity first inferred by Gussow (1953) and subsequently by Jutras et al. (2007) farther north in southern Québec and northern-most New Brunswick. Limited spore recovery from these red beds also has hindered biostratigraphic subdivision (St. Peter and Johnson 2009).

3 A recent potash exploration program by the Potash Corporation of Saskatchewan Inc. (PotashCorp), near Sussex, now has resulted in the collection of over 5 km of drill-core. Sedimentary logging of these cores, together with bulk-rock Inductively Coupled Plasma – Mass Spectrometry (ICPMS) analyses of 131 samples from three of the cores, allows for new insights into the red bed succession, the informal stratigraphy, and supporting evidence for the previously identified unconformity. The results further confirm the utility of whole-rock chemostratigraphy.

4 The > 3.5 km-thick Devonian-Carboniferous stratigraphic succession in southern New Brunswick overlies terranes (southwards, Ganderia, Avalonia, and, in Nova Scotia, Meguma) that accreted to Laurentia in the Silurian-Devonian (Fig. 1). However, in this region these terranes were still subject to later relative movements accompanying oblique collision of Gondwana to Laurentia (Alleghanian orogeny, Murphy et al. 2011; Waldron et al. 2015). This provided a major transtensional-transpressional structural control on the younger stratigraphy. Initially, the result was the development of small-scale sedimentary pull-apart basins, such as the Moncton Basin (Craggs et al. 2017).

5 The three tectono-sedimentary megacycles recognised in the New Brunswick Devonian-Carboniferous succession (Fig. 2; St. Peter 1993) are separated by interbasinal angular unconformities related to specific tectonic episodes (Murphy et al. 2011). More localized intrabasinal unconformities are identified within individual basins. The first megacycle, which culminated in a late Tournaisian to early Visean inversion, consists of strata assigned to the Horton and Sussex groups. The second megacycle contains strata assigned to the Windsor and Mabou groups. The subsequent late Visean to Serpukhovian inversion (Wilson and White 2006; Murphy et al. 2011), is related to the near-continent-wide Mississippian-Pennsylvanian unconformity, and in eastern Canada is signified by a major episode of transpression and uplift associated with the Minas Fault Zone (Fig. 1a; Murphy et al. 2011). Early Pennsylvanian deposition (Cumber-land Group) again appears to have been initiated in areas of transtension, and rapidly subject to salt tectonism associated with deformation of buried Windsor Group evaporites (Waldron et al. 2013; Craggs et al. 2017). However, in the later Pennsylvanian, thermal relaxation caused regional cooling and subsidence, leading to the development of an enlarged sedimentary basin, which buried most of the rapidly deposited, locally sediment-sourced, and deformed pull-apart basins. Therefore, in contrast to the first two megacycles, the third evolved to also consist of a blanket of more mature, generally more slowly accumulated sediment (Pictou Group) in a more regional terrestrial drainage sys-tem (Gibling et al. 1992).

6 The focus of this paper is on red beds currently included in the upper Mississippian 2^nd megacycle (Fig. 2). Within this megacycle, carbonate and evaporite rocks are interpreted to have been deposited following marine transgressions, and red beds to have followed regional uplift or marine regression (Giles 1981, 2009; Wilson 2005). Various formations are combined into the Windsor Group, the lower and upper contacts of which are placed at the base and top of the strati-graphically lowest and highest carbonate or evaporite beds encountered in a particular area (Giles 1981; McCutcheon 1981). In the western Moncton Basin, the uppermost Windsor Group comprises the mined potash and halite (Cassidy Lake Formation) and overlying anhydrite and halite (Clover Hill Formation, McCutcheon 1981; Wilson et al. 2006).

7 In Nova Scotia, additional cycles of red beds, carbonate rocks, and sulfate rocks are reported overlying Cassidy Lake equivalent strata (e.g., Giles 2009); in New Brunswick a red bed succession up to 1 km thick overlies the Cassidy Lake Formation (Webb 2009). Originally considered part of the Gypsiferous Formation (Lyell 1843) or part of Division 8 of the Coal Measures (Logan 1845), and later the Bonaventure Formation (Logan 1864), these red beds in the far southeast of New Brunswick were formally assigned to the Hopewell Group and divided into three units by Norman (1941a, b) and Gussow (1953). A lower, red mudstone-dominated unit (Maringouin Formation, now Middleborough Formation, Jutras et al. 2016), is overlain by interbedded red and grey sandstone, red and grey mudstone, and minor calcareous mudstone-pebble and coalified plant-fragment conglomer-ate, with Cu-mineralized horizons at the base of the unit (Shepody Formation). The uppermost red beds are interbedded conglomerate, sandstone, and mudstone (Enragé Formation). Toward the basin margins, a localized conglomeratic unit interfingers with and progressively overlies Windsor Group strata (Hopewell Cape Formation). Despite the ear-lier names noted above, for stability of nomenclature most authors (e.g., Wilson et al. 2006; Force and Barr 2006; St. Pe-ter and Johnson 2009; Utting et al. 2010; Murphy et al. 2011; Allen et al. 2013; Waldron et al. 2013; Hayward et al. 2014; Holt et al. 2014; Stimson et al. 2016; Craggs et al. 2017) in-clude these and laterally equivalent red bed units sitting on Windsor Group strata in the Mabou Group of Belt (1964, 1965). Overlying plant-bearing grey sandstone, mudstone, and thin coal beds are assigned to the Boss Point Formation (Cumberland Group, Pennsylvanian).

8 In the Sussex region, the informal units of Anderle et al. (1979) have been used for the post-Windsor red beds, namely: grey and reddish brown sandstone, siltstone, and minor conglomerate for the Poodiac formation; greyish red conglomerate, and minor sandstone and siltstone for the Wanamaker formation; and red siltstone, and fine sandstone for the Scoodic Brook formation (note that, following the North American Commission on Stratigraphic Nomenclature, 1983, informal names do not warrant capitalization of “formation”).

9 There has been much debate regarding the nature of the contacts associated with the post-Windsor red beds. Gussow (1953) proposed that a “mid-Hopewell” unconformity separating the Shepody and Enragé formations was equivalent to the Mississippian-Pennsylvanian boundary. As evidence, in a seismic-based cross-section (adjacent to the current study area) he interpreted Shepody-involved thrust faulting to be regionally truncated by overlying Enragé Formation. This latter formation, and its equivalent Claremont Formation in Nova Scotia, also have been assigned to the overlying (i.e., 3^rd megacycle) Pennsylvanian Cumberland Group by Ryan and Boehner (1994). More recent seismic interpretations identify “a local intra-Mabou unconformity” adjacent to a thrust-related triangle zone, with Gussow’s regional Mississippian-Pennsylvanian boundary unconformity picked at the Enragé – Boss Point contact (Wilson and White 2006). This latter contact for the Mississippian-Pennsylvanian boundary is also supported by the palynological work of Utting et al. (2010).

10 In Québec, Jutras et al. (2001) noted a channelized disconformable contact between underlying red clastic rocks of the type Bonaventure Formation and the overlying plant-bearing grey clastic rocks of the Pointe Sawyer Formation. Although lacking biostratigraphic control, the underlying red beds were considered time equivalent to the upper Windsor Group strata of Nova Scotia and named the Percé Group by Jutras and Prichonnet (2005). As noted above, this splitting of the red beds has not been adopted by other workers in New Brunswick and Nova Scotia.

11 Drill cores (2.5 inch, or 63.5 mm, diameter) from the 2002, 2005, and 2006 PCS exploration program were selected. From southwest to northeast of the study area (Fig. 1c) these holes are named: PCS-02-01, PCS-05-04, PCS-02-04, PCS-06-01, PCS-02-02, PCS-05-02, PCS-02-03, PCS-02-05, and PCS-04-01. Cores from PCS-05-01 and PCS-05-03 were excluded from detailed lithofacies analysis because they are highly fractured. Logging of the approximately 4 km of core was undertaken at the centimetre scale from the depth where coring began in the red beds, down to the contact with underlying evaporite or carbonate; study of Windsor Group strata did not form part of this project. Lithofacies analysis was based on the classification scheme of Miall (1978, 1996).

12 A small biostratigraphic study was attempted on material from grey siltstone beds but no palynologic components were identified that could narrow down the time of deposition beyond what is already known and stated above (G. Dolby, personal communication, 2013).

13 Several of the logged boreholes were also selected for geochemical sampling. For PCS-02-05, 59 samples were collected systematically at regular (~10 m) intervals. For boreholes PCS-05-04 (n = 44) and PCS-02-01 (n = 30), samples were collected by random number generation (converted to borehole depth) using Microsoft Excel to achieve more unbiased results. In all cases, to minimize variation resulting simply from grain size factors (Pearce et al. 1999), materials were collected from the finest grained beds within 1 m of the designated sampling depth.

14 Following Pe-Piper et al. (2008), sample preparation and analyses were performed by Activation Laboratories Ltd. (Actlabs) in accordance with the International Organization for Standardization (ISO) 17025. Preparation procedures are detailed on the Actlabs website (Activation Laboratories Ltd. 2014). Powdered and fused samples were analyzed by Perkin Elmer Sciex ELAN 6000, 6100 or 9000 ICP/MS, and samples from each core were randomized prior to analysis. Three blanks and five controls (three before the sample group and two after) also were analyzed per group of samples. Control samples included standards NIST694, DNC-1, and BIR-1a for major elements, and OREAS 100a and 101a, NCS DC 86312, 70009, and 70014 for minor and trace elements. The relative standard deviation from replicate analyses of the standard is <5% for major elements and <10% for minor/trace elements. The uncertainty associated with various determinations is ± 100 % at the detection limit (DL), ± 15 % at x10 DL and ± 5 % at x100 DL. Duplicates were fused and analyzed every 15 samples with instrument recalibration after every 40 samples.

15 Analyses of the different cores were run several months apart. Although the same standards were run with each analysis, and quality control checks indicate consistent readings, during statistical analyses we follow Tuttle et al. (1983) and do not combine data from different cores into single datasets.

16 In whole-rock geochemical data, variables are non-negative and compositional in nature; values are recorded as a proportion of the whole, i.e., ppm or percent (%). This constrained sample space, or the simplex, leads to what is known as the constant sum problem, which together with an often non-normal distribution of the data, makes basic statistical analyses (e.g., variance, correlation, regression and its derivatives) inapplicable (Pearson 1896; Chayes 1960; Butler 1979; Davis 2002). It also means that the compositional variables are, in fact, ratios whose denominators are rarely the sum of all the constituents available to be measured (Davis 2002). To overcome this problem, Aitchison (1986) demonstrated that centred log ratio (CLR) data transformation allows for the use of standard statistical methods on such geochemical data (Can Mert et al. 2016). This transformation is a simple process done by taking the natural log of each quotient where each compositional variable is divided by the geometric mean of the composition. Elements with more than one recorded value below detection limit remain problematic. For elements with only one such value (cobalt, zinc, germanium, tin, thallium), an artificial value (v) was assigned for statistical purposes, where v = x – (x/10) and x = detection limit (Davis 2002). Elements with more than one value below detection limit were excluded from statistical analyses. The raw data (in ppm) are provided in Appendix A.

17 Sedimentological analysis of red bed cores taken from most of the recent vertical PCS exploration boreholes (Figs. 1c, 3, 4) resulted in the identification of 11 major lithofacies (Table 1). Lithofacies are mainly reddish-brown, pale brown, brownish-grey, and brown; ferruginous or calcareous; and mainly horizontally stratified. A unique bluish- or brownish-grey sandstone interval containing rip-up clasts and/or plant fragments occurs in the lower parts of most boreholes. The summary sedimentological logs (Fig. 3) indicate that fine-grained and/or sandstone-dominated facies broadly coarsen upward to gravel facies, with the amount of coarse-grained sediment also increasing toward the northeastern part of the study area. Accordingly, four intervals, the lower red sandstone-siltstone interval, bluish-grey sandstone interval, upper red sandstone-siltstone interval, and red polymictic conglomerate interval, have been identified. They are briefly reviewed in terms of their lithofacies associations. Also for brevity, since limited information is preserved in core (e.g., lateral facies relationships and larger scale geometries are not present to narrow down an interpretation), a plethora of possible conditions exist under which these various sedimentary structures are deposited, and a comprehensive discussion of all the possible interpretations is not feasible: only our favoured interpretations are presented.

28 The grey sandstone interval is a distinct stratigraphic marker in the study area, the base of which, as interpreted, marks a shift in depositional conditions. The bulk geochemical data collected (from PCS-02-01, PCS-05-04, and PCS-02-05) can be used to statistically test whether the base of the grey sandstone interval is also a distinct chemostratigraphic horizon. Simply, for comparison of variances and mean values, the null hypothesis can be tested against its opposite:

where below the base grey sandstone interval is considered as ‘lower interval’, and the base grey sandstone interval and above are considered as ‘upper interval’. Significance was set at α= 0.05 in all analyses. Differences in correlation coefficients and simple regression equations (with SiO₂ as the explanatory variable) provide additional comparisons.

29 Geochemical data can be presented in different ways, including simple plots of specific elements against depth or location, and scatter plots (Harker diagrams and discriminant plots) of two elements or two computed groups of elements, e.g., to identify provenance. In the selected plots presented herein (Figs. 5 to 9), data from the lower interval are coloured blue, and data from above the base of the grey sandstone interval are coloured red.

42 The general lithofacies distribution in the studied boreholes suggests an alluvial fan to floodplain model, with the proximal fan located toward the northeast (Figs. 3, 10). Over time, the conglomerate units were able to prograde further toward the southwest. Of particular interest, however, is what we demarcate as the “bluish-grey interval”, where exists a distinct combination of metre-scale bluish-grey beds of sandstone (non-red zones exist elsewhere but not at this scale), red or grey mudstone pebbles (i.e., rip-up clasts – also present, but to a lesser extent, at other stratigraphic levels) and, unique to this interval, plant detritus. As previously noted, preservation of plant fragments and patchy grey colouration indicate a more prevalent reducing environment. The presence of rip-ups suggests associated incision and reworking by channelization. This could be related to a rise in the water table (the terrestrial proxy for ‘base level’), either related to relative sea- or lake level rise or simply a wetter climate resulting in more perennial discharge. The geochemical data above and below the base of the bluish-grey interval in addition indicate significant differences in the elemental compositions that likely reflect more varied provenance in the younger strata. This change could be due to gradual unroofing of additional rock types in the hinterland drainage basin.

43 In seismic profiles that cross the study area (Wilson and White 2005; fig 3b in Keighley 2008) a mid-Mabou angular unconformity, passing south into a (dis-) conformity can be picked, ~300–400 m above top Windsor, to the north of the subsurface Windsor-salt diapir (Fig. 1c). The Mabou Group is everywhere poorly imaged south of the diapir in the study area, but the bluish-grey interval might sit on the same unconformity south of the diapir. If so, rip-up clasts may be the result of vertical, rather than lateral, incision due to relative base-level fall; the reducing environment was potentially the result of base-level rise during subsequent infill of the incised valley. The lack of any distinct change in the dips of beds in PCS cores from above and below the base of the grey-coloured interval favours a disconformable contact, but the limitations of assessing minor changes in bedding from core material cannot rule out an angular boundary. Elemental geochemistry indicates a more varied suite of detrital minerals above the base grey-coloured interval (i.e., above the unconformity). This is indicated in the provenance plots and the goodness of fit to regression lines of Si:Al and K:Na, and statistically confirmed in f-tests of variance of Nb:Ta and Ti:Cr ratios between upper and lower intervals (Figs. 5 to 9). If tectonic uplift was involved in producing the unconformity, more varied source rocks could have been unroofed, weathered, and eroded from the hinterland. Alternatively, stream capture producing a larger hinterland exposing more varied lithologies may have resulted from uplift, or simply wetter climatic episodes, related to renewed relative base level rise following the development of the unconformity (Fig. 10).

44 Two different lithostratigraphic interpretations of the succession are possible. First, the grey sandstone interval, not previously recognized in outcrop in the Sussex area, may be considered equivalent to the Shepody Formation in the far southeast of New Brunswick. The lower red sandstone-siltstone interval would then be equivalent to the Middleborough Formation (syn. Maringouin Formation), and the upper red beds possibly equivalent to the Enragé Formation. In the type sections (southeastern New Brunswick), the thicknesses of the Maringouin, Shepody, and Enragé formations are ~496 to ~700 m, ~200 to ~330 m, and ~100 to ~150 m, respectively (St. Peter and Johnson 2009). If this interpretation is accepted, it would be the first indication in the Moncton Basin of a base-Shepody Formation disconformity, as Gussow (1953) previously identified an unconformity that he associated with the top of the formation. However, in contrast to the type sections, thicknesses of the equivalent ‘formations’ are quite different in the study area: ~220 to ~97 m, ~75 to ~7 m, and ~300 to ~10 m, respectively.

45 Alternatively, where the Windsor Group is more extensively exposed in Nova Scotia, five transgressive-regressive cycles are delineated (Giles 2009). Interfingering red bed intervals can be more than 250 m thick (Ryan and Boehner 1994), and a major unconformity is identified in the Asbian. Contrastingly, in the Moncton Basin, no late-Asbian (cycle II, e.g., Lime-kiln Brook Formation) and post-Asbian (cycle III to IV) Windsor Group carbonates or evaporites are known (St. Peter and Johnson 2009): the Lime-kiln Brook Formation is present only in the far SE part of New Brunswick in the Cumberland Basin (Fig. 2; Craggs et al. 2017). If the final marine regression of the Windsor Group sea in the Moncton Basin took place before the development of the Asbian unconformity, then an alternative lithostratigraphy for the Sussex area could be that the lower red bed unit, although by definition Mabou Group (lying above the uppermost preserved limestone), may be a Windsor Group cycle I landward equivalent. The capping grey sandstone and upper red beds would be cycle II transgression and regression equivalent or younger (Fig. 10), and strata equivalent to the Shepody Formation would, if ever deposited, overlie the studied section. Red beds disconformably underlying grey and plant-bearing sandstones have also been considered Windsor Group equivalent in Québec, where they have been assigned to the Percé Group (Jutras et al. 2001; Jutras and Prichonnet 2005). Although the latter interpretation is favoured, a final decision as to the most appropriate lithostratigraphy is still considered premature, because the unconformity has still not been mapped (and confirmed) in outcrop.

46 One of the main constraints on previous studies of the red bed succession assigned to the Mabou Group of New Brunswick has been limited data due to poor outcrop exposure. However, access to nearly 5 km of recent drill cores from 11 boreholes has allowed sedimentological logging of several hundred metres of a continuous red bed succession. The succession can be divided into four intervals, i.e., the lower red sandstone-siltstone interval, bluish-grey sandstone interval, upper red sandstone-siltstone interval, and red polymictic conglomerate interval. The lower three intervals are interpreted as braidplain or distal alluvial fan deposits. The conglomerate interval indicates predominately debris flow alluvial fan deposits. The bluish-grey sandstone interval is associated with relative base level change that may simply be related to climate change or possibly to the development of a disconformity at its base.

47 A disconformity within post-Windsor Group red beds is supported by whole-rock geochemistry of 131 samples from three of the drill holes. Elemental variations observed through analysis of major element oxides and trace elements suggest a more varied or mixed provenance above the grey-coloured sandstone interval. For example, Nb:Ta and Ti:Cr ratios indicate a significant difference in the variance between upper and lower intervals.