Abstract

The widely held view that Pennsylvanian cyclothems formed in response to Milankovitch-controlled, glacio-eustatic, sea-level oscillations lacks unambiguous quantitative support and is challenged by models that are based on climate-controlled precipitation-driven changes in depositional style. This study shows that cyclothem successions do in fact contain a clear record of Milankovitch-controlled oscillating sea level, but that it is prerequisite that besides cyclothem thickness, cyclothem composition is taken into account. A simple subdivision of cyclothems into subaqueous and subaerial facies is sufficient to reveal the signal, provided that sufficiently long and complete successions are studied.

Two Duckmantian–Bolsovian (Westphalian B–C) successions were studied—one from a high-accommodation setting in the Netherlands and another from a medium-accommodation setting in Kentucky in the United States. The Dutch record comprises an exceptional, 1728-m-long, continuously cored interval, and it shows a distinct twofold cyclicity in the subaerial-facies ratio of subsequent cyclothems at wavelengths of ∼256 m and ∼59 m, which is confirmed by power-spectral analysis. The signal is not present in the Kentucky succession due to subsidence-controlled low preservation of only one out of three to four cyclothems, and that explains why many cyclothem studies have yielded inconclusive results.

Recent U/Pb ages indicate that the 256 m cycle represents ∼395 k.y., which matches with long eccentricity (413 k.y.). This then gives a 95 k.y. duration for the 59 m cycle (short eccentricity). Individual cyclothems in the high-accommodation Dutch succession are mostly between 5 and 35 m thick, which points to a sub-eccentricity duration (mean 21 k.y.). The highly variable thickness may be due to interference of precession-, obliquity-, and eccentricity-driven sea-level fluctuations or alternatively to autocyclic or climate-controlled variations in sediment supply. Integration of the results with U/Pb calibrated radiometric ages for “tonstein” ash layers from North America and Europe allowed refinements of the chronology of the main Westphalian (Moskovian–Bashkirian) coal interval; these refinements are consistent across Euramerica.

An analysis of cumulative coal-bed thickness further indicates that terrestrial-carbon (coal) storage patterns are comparable in the two remote areas: in the Netherlands ∼5 m coal per m.y. during the Langsettian (Westphalian A) and increasing abruptly to ∼20 m/m.y. at the start of the Duckmantian substage (Westphalian B). In Kentucky, storage rates were lower, but when standardized to Dutch subsidence, the pattern is identical. This suggests that burial of terrestrial carbon during the Late Paleozoic Ice Age was globally controlled and possibly very predictable.

INTRODUCTION

Pennsylvanian cyclothems, deposited in paleoequatorial Euramerican basins during the late Paleozoic ice-house period, are a classic example of sedimentation controlled by glacio-eustasy, with sea-level fluctuations driven by the waxing and waning of Southern Hemisphere ice caps (Veevers and Powell, 1987). The concept of sea-level control emerged in the first half of the twentieth century in publications by Udden (1912), Weller (1930), and in particular Wanless and Shepard (1936), who linked the cyclic character of coal-bearing Pennsylvanian deposits with growing evidence for widespread late Paleozoic glaciation. The case was strengthened during the latter part of the twentieth century with a large body of studies dealing with various aspects of the cyclothem formation, including possible Milankovitch control (Busch and Rollins, 1984; Heckel, 1986; Veevers and Powell, 1987; Klein and Willard, 1989; Davies et al., 1992; Maynard and Leeder, 1992; Aitken and Flint, 1995; Greb et al., 2008; Heckel, 2008).

Although the influence of glacio-eustatic sea-level fluctuations on deposition during the Pennsylvanian became accepted in the general sense, skepticism has remained, not in the least because the resolution of data sets and the precision of the absolute time scale have proven insufficient to unambiguously demonstrate a causal and temporal link (Algeo and Wilkinson, 1988; Klein, 1990; Wilkinson et al., 2003). An alternative mechanism was proposed by Cecil (1990), who argued that—in sync with eccentricity-driven sea-level fluctuations—the formation of Pennsylvanian cyclothems was primarily controlled by paleoclimate, with precipitation cycles triggering the alternation of coal and clastics. This model has gained popularity recently (Cecil et al., 2003; Eros et al. 2012; Rosenau et al., 2013; Cecil et al., 2014; DiMichele, 2014). It explains the formation of cyclothems still within the framework of glacio-eustasy and sequence stratigraphy (Cecil et al., 2014; Eros et al., 2012), but the model differs in the mechanism by which the glacio-eustatic signal is translated to the sedimentary system (driven by sediment supply rather than accommodation) and in the timing of deposition of the different facies (e.g., peat accumulation during lowstand of sea level rather than transgression). This study focuses on the role of glacio-eustasy in cyclothem formation, and we test whether the typical “cyclothemic” alternation of marine and non-marine sediments can be unambiguously explained in terms of Milankovitch control.

Sea-level fluctuations are also believed to have triggered the accumulation of peat layers (coal) over vast areas during early transgression (Heckel, 1990; Flint et al., 1995; Bohacs and Sutter, 1997) or under conditions of high precipitation during sea-level lowstand (Cecil et al., 2014; DiMichele, 2014). Pennsylvanian coal beds constitute the largest reservoir of terrestrial carbon on the globe (Berner, 2003). Hence, when linked with volcanic-ash dates, thick cyclothem successions may not only hold the key to detailed reconstruction of sea-level fluctuations during a major ice-house period and to the fine-tuning of late Paleozoic chronology but also to accurate estimates of global carbon fixation rates in coal swamps.

Although estimated cyclothem periods are in the Milankovitch range (Heckel, 1986; Maynard and Leeder, 1992) and cyclothem bundling patterns match Milankovitch-cycle ratios (Busch and Rollins, 1984; Heckel, 1986), orbital control has not been demonstrated unambiguously and lacks a strong quantitative support (Algeo and Wilkinson, 1988; Wilkinson et al., 2003; Meyers et al., 2008). Difficulties in matching data with the Milankovitch model include poor time control (Klein, 1990), incomplete successions (De Boer, 1991; Hinnov, 2013), nonlinear time-sediment relations (Algeo and Wilkinson, 1988; Hinnov, 2013), and the interference of cycles (De Boer, 1991). Therefore, the extraction of accurate sea-level records, which could contribute to the understanding of ice-sheet dynamics, terrestrial carbon storage, and other aspects of paleoclimate, remains a challenge.

In older studies, Pennsylvanian cycle periods are variably linked with precession (17 k.y.), obliquity (35 k.y.), and short and long eccentricity (95–413 k.y.) (Heckel, 1986; Klein, 1990; De Boer, 1991; Maynard and Leeder, 1992; Goldhammer et al., 1994), but the low resolution of the Pennsylvanian time scale has hampered detailed cycle determinations (Klein, 1990). Various recent studies, however, have indicated that cyclothem deposition during the Pennsylvanian was primarily forced by short eccentricity (Greb et al., 2008; Eros et al., 2012; Waters and Condon, 2012). However, this alleged Pennsylvanian dominance of short eccentricity is not in line with Quaternary insolation models. These predict alternating eccentricity and obliquity control, as well as some influence of precession (Ruddiman, 2006), which is reflected in δ18O records for much of the Quaternary (Hays et al., 1976; Raymo et al., 1990).

Most cyclothem studies have been performed in low to medium subsidence areas in the USA and the UK, where Pennsylvanian successions are relatively thin (∼100s m to ∼1 km). In high-subsidence areas such as Nova Scotia, continental Europe, as well as a number depocenters in the southern and central parts of the Appalachian Basin, Mid-Pennsylvanian successions are many kilometers thick (Drozdzewski, 1993; Falcon-Lang, 2004; Greb et al., 2008), and cyclothems are so numerous that they must represent shorter time intervals to be consistent with age data. Hence, the interpreted dominance of eccentricity cycles in low to medium subsidence areas, where successions are much thinner, could well be preservation driven. Chesnut (1997) noted that cyclothems with a ∼100 k.y. duration in the central Appalachian Basin grade into cyclothems with a 400 k.y. duration toward areas of lower subsidence, indicating that multiple depositional hiatuses and/or erosion surfaces may characterize areas of low accommodation.

In this study, a Langsettian–Bolsovian (Westphalian A–C) cyclothem succession was analyzed from a high subsidence area in the Netherlands, which is ∼1728 m thick and covers ∼145 cyclothems. The interval was cored completely, allowing analysis of the gradual compositional variation of cyclothems over time. This resulted in a much higher resolution cyclicity interpretation than based on cyclothem thickness alone. The succession was compared with an overlapping cyclothem succession from a medium-accommodation setting in Kentucky (USA) to see how subsidence rate affected deposition and the preservation of cyclothems. The results were used to refine Middle Pennsylvanian chronology and to determine coal thickness per unit time for coal basins.

PERIODICITY IN CYCLOTHEM RECORDS

Thickness and average cycle length have been the primary data source in cyclothem analysis. Thus far this has not yielded conclusive evidence in terms of cycle periods. This may be related to the above-mentioned problems, such as poor time control, but may be intrinsic to the method as well. Counterintuitively, cyclothem boundaries are not typically created at sea-level highstand or lowstand. For example, it may take many thousands of years before rising sea level causes flooding of the land and before deposition of marine shales on top of a previous cycle starts. This is illustrated by the example in Figure 1, which is based on the simple, theoretical depositional model of Jervey (1988). It assumes that subaqueous facies are deposited when base level is above the land surface, followed by non-deposition (or local erosion) when it falls below the land surface, followed by subaerial deposition when rising base level meets the land surface again and as long as it is balanced by sediment supply.

The synthetic sea-level curve in Figure 1 combines precession (17 k.y.), obliquity (35 k.y.), and eccentricity (95 and 413 k.y.) cycles (Pennsylvanian cycle periods after Berger and Loutre, 1994) with relative amplitudes of 0.5, 1, 1, and 0.5. The predicted cyclothem succession displayed in the figure shows the thickness as well as the composition of the cyclothems in terms of subaqueous and subaerial facies.

The individual cyclothems have a predicted duration between 9 and 82 k.y. (Fig. 1B). None of the basic input cycles stands out, but with ∼50% of the cycles in the 29 to 39 k.y. range, the record suggests obliquity dominance; although the input sea-level amplitudes for obliquity and short eccentricity were equal. Short eccentricity is represented as trimmed cycles shorter than 82 k.y., and long eccentricity is faintly present as a long-term fluctuation of cyclothem thickness. A comparison with the eccentricity components of the input sea-level signal shows that the cyclothem thickness fluctuations are locally in phase with sea-level change but frequently are out of phase (Fig. 1B). From the model, it appears that input sea level can be more accurately reconstructed on the basis of the fluctuation of the percentage of subaqueous facies in subsequent cyclothems (Fig. 1C), which is referred to here as the “subaqueous-facies ratio.” In that signal, both the 95 k.y. and the 413 k.y. stand out prominently, and a comparison with input sea level shows that peaks line up well with the eccentricity signal (Fig. 1C).

The above model may explain why cyclothem-thickness records so far have not yielded conclusive evidence of Milankovitch control, and it indicates that results may be better if the composition of cyclothems is taken into account. In addition to the subaqueous-facies ratio, the fluctuation of the percentage of coal within the subaerial parts of cyclothems was analyzed.

THE PENNSYLVANIAN OF EURAMERICA

The Dutch wells are from the Northwest European thermal-sag basin located to the north of the Variscan orogenic chain (Fig. 2). It contains an up to 3–4-km-thick Pennsylvanian succession with only gradual thickness variations over relatively great distances throughout most of the basin (Drozdzewski, 1993). The studied Kentucky interval is from the central Appalachian Basin, which during the Pennsylvanian was an E/W-aligned, elongate foreland basin. It was at times separated from the larger Late Pennsylvanian Midcontinent Sea that lay to its north (Algeo and Heckel, 2008) by the subdued highs of the Cincinnati-Findlay-Algonquin arch system (Tankard, 1986). The basin contains an up to 1.5-km-thick Pennsylvanian interval that rapidly thins westward. Both basins were located near the paleoequator and were ∼3000 km apart during the Late Carboniferous (Blakey, 2014).

In both areas, the coal-bearing successions span the Langsettian to Bolsovian substages (Westphalian A–C; Fig. 3) and consist of stacked cyclothems, mostly 10–25 m thick, of alternating (marginal) marine shales and coal-bearing lower and upper delta-plain deposits. The lower ∼2 km of the Dutch interval is shale dominated; 10–20-m-thick fluvial sandstone bodies only become prominent in the Upper Bolsovian. In Kentucky, such sandstone bodies are present throughout the succession and constitute the basal parts of cyclothems as defined by, for example, Weller (1930), Klein and Willard (1989), and Aitken and Flint (1995). These commonly overlie regionally extensive erosion surfaces that cut down into underlying strata.

In both basins, fossiliferous marine shales are present at the base of the Duckmantian (Vanderbeckei/Betsie) and Bolsovian (Aegiranum/Magoffin), and at ∼40% below the top of the Duckmantian (Maltby/Kendrick) (Fig. 3). These “marine bands” are attributed to highstands of global sea level and can be correlated between the basins (Riley and Turner, 1995).

A facies example from a road cut along highway US 119 in eastern Kentucky is presented in Figure 4, which shows a fluvial-sandstone–dominated section in its upper part and a marine shale at the base (Kendrick Shale, European equivalent: Maltby/Domina M.B.).

ANALYSIS

More than 2.5 km of continuous core from the Pennsylvanian in wells RLO-1, HGV-1, JOP-1, and KPK-1 from the Netherlands was studied (Fig. 2). A 1728-m-long composite section was built spanning the Late Langsettian–Late Bolsovian (Van de Laar and Fermont, 1989). Well KPK-1 was drilled ∼140 km south of the other wells, but regional thickness differences are small (Van de Laar and Fermont, 1989).

For comparison, a composite sedimentary section was recorded from large, fresh road cuts along highway US 119 in eastern Kentucky (example shown in Fig. 4). The section spans the Duckmantian (Westphalian B), which is only 320 m thick due to a more limited stratigraphic coverage and a lower subsidence rate (∼35%).

Cyclothem successions were recorded as alternations of subaqueous deposits (laminated and/or bioturbated shales and mouthbar sands), and subaerial deposits (coals, rooted shales, and fluvial sands) and cyclothem boundaries were placed at the base of the subaqueous interval (cf. Galloway, 1989). The Dutch succession is little affected by fluvial incision, possibly as a result of much higher subsidence rates. In the few cases where the presence of thick fluvial sandstone bodies is associated with unconventionally thick cyclothems (e.g., top half of road cut in Fig. 4), this was interpreted as the result of erosion of the subaqueous part of a cyclothem due to fluvial processes during lowstand (cf. Davies et al., 1992; Aitken and Flint, 1995). In such cases, a cyclothem boundary was placed at the midpoint of the composite cycle. In the Kentucky succession, thick fluvial sandstones are so numerous that this was not considered feasible, and fluvial sands were included in the subaerial part of cyclothems resulting in a less accurate record.

The Dutch wells were merged into a composite succession based on palynology by Van de Laar and Fermont (1989), and the correlation was fine-tuned using the cyclothem data. The Kentucky section was matched with the Dutch composite using the marine bands at the base of the Duckmantian and Bolsovian as anchor points for correlation (Riley and Turner, 1995). For each cyclothem, the subaqueous-facies ratio and the percentage of coal in its subaerial part were calculated.

The presence of cyclic signatures in the records was evaluated using Blackman-Tukey spectral analysis, making use of AnalySeries 1.1. software (Paillard et al., 1996). This was followed by a visual inspection at the detected frequencies using simple sine waves. Calculation of cycle periods was based on an integration of Ar/Ar and U/Pb radiometric dates.

RESULTS

Results for the Dutch wells are presented in Figure 5A. Note that the subaqueous-facies ratio is plotted in mirror image, i.e., as a subaerial-facies ratio, giving a better visual representation. The original correlation based on palynology by Van de Laar and Fermont (1989) is confirmed by the data presented here and needed only a minor adjustment. Wells HGV-1 and JOP-1 overlap with KPK-1, and after thickness adjustments of 85% (JOP-1) and 95% (HGV-1) correcting for differential subsidence, a good match was achieved, in particular for the coal percentage (Fig. 5A, third graph). A thickness adjustment to 90% (average of the nearby wells HGV-1 and JOP-1) was applied to well RLO-1 to correct for slightly higher subsidence compared to the more southerly well KPK-1. From RLO-1 and KPK-1 an Upper Langsettian to Upper Bolsovian composite of 145 cyclothems was constructed (Figs. 5B and 5C).

The Dutch succession was matched with the more condensed succession from Kentucky according to the interbasinal correlations of the marine bands at the base and top of the Duckmantian proposed by Riley and Turner (1995). These correlations are based on comparable conodont, ammonoid, and palynological content. Matching required a 2.8 stretching factor. The overlapping Duckmantian interval, which consists of 62 cycles in the Netherlands, is represented by only 16 cycles in Kentucky. Visual comparison of the two successions shows only moderate correlation, which is probably partly due to the different number of cycles and the stretching correction. The cycle thickness patterns do not show an obvious match (Fig. 5C, first graph). The subaerial-facies ratio data (second graph), however, match well; the number of measurements is obviously lower, but the general trends are comparable. Trends in the coal-percentage record are not a perfect match but still show a fair degree of correlation (third graph). It is further noted that the major coal peaks line up well. The fact that the composition of the cyclothems, in particular the subaerial-facies ratio, shows a much better match is in line with the model predictions presented earlier (Fig. 1). Although the results do not prove the ammonoid, conodont, and palynology-based interbasinal correlations proposed by Riley and Turner (1995), the similarity in trends for the subaerial-facies ratio and the coal percentage do indeed support it.

Cyclothem thickness in the Dutch interval varies strongly, with cycles distributed evenly between 5 and 20 m, and occasional cycles between 20 and 36 m. The average cycle is 12.5 m thick. The Kentucky cycles are evenly distributed between 3 and 27 m with occasional cycles up to 44 m and an average of 17.3 m. This greater mean thickness of the Kentucky cycles is attributed to less compaction due to a much higher sandstone percentage and to the undetected merging of some of the cycles due to fluvial erosion and the formation of condensed paleosol intervals during lowstands (Aitken and Flint, 1995).

CYCLICITY INTERPRETATION

Blackman-Tukey power spectra were calculated for the subaerial-facies ratio and the coal-percentage data sets for the extensive Dutch data set (Fig. 6). The spectrum for the subaerial-facies ratio shows two pronounced cycles with wavelengths of 256 m and 59 m and a subordinate, more subdued peak with a wavelength of 146 m. The 256 m cycle observed in the subaerial-facies ratio is repeated as a well-defined peak in the coal-percentage spectrum, although at a slightly reduced thickness of 245 m. Higher-frequency peaks in the coal-percentage data set are poorly resolved; minor peaks were detected at periods of 193 m, 74 m, and 46 m.

Sine-wave overlays were used to visualize the sedimentary cyclicity in the Dutch succession. Because the results for the two data sets were very comparable but more accurate and reliable for the subaerial-facies ratio, the wavelengths determined for that specific spectrum were used to construct the sine waves (256 m and 59 m).

The 256 m cycle has a consistent, strong visual presence in the subaerial-facies ratio as a gradual increase from ∼10% to ∼80% subaerial-facies per cycle and a gradual decrease back to ∼10% over ∼4–5 higher-frequency cycles in the subaerial-facies ratio (Fig. 7A, top graph). In the coal-percentage record, it is not well developed; although in parts of the data set, peaks line up moderately well with the sine-wave overlay (Fig. 7A, bottom graph).

The other well-developed peak (59 m) is also prominently visual in the subaerial-facies ratio as the highest frequency fluctuation between a low and high ratio for subaerial facies (Fig. 7B, top graph). It represents a gradual increase and decrease through a number of subsequent cyclothems. In the coal-percentage record, the separation of subsequent peaks is of the same order, and many peaks line up well with the overlay, although a bit offset in places. Again the cyclicity is less pronounced than in the subaerial-facies ratio data set (Fig. 7B, bottom graph). The subdued peaks representing wavelengths of 146 m (subaerial-facies ratio) and 114 m (coal percentage) could not be recognized visually in the data sets.

Absolute cycle periods were determined using recent U/Pb age determinations of volcanic-ash layers in the Duckmantian Fire Clay coal (314.6 Ma) and in the Langsettian Upper Banner coal (316.1 Ma; Lyons et al., 2006). The position of the Fire Clay coal is known (Fig. 7C), and the approximate position of the Upper Banner coal bed was extrapolated from a nearby gas well (Rice et al., 1987) to slightly below the base of the Dutch composite succession. The ∼1.5-m.y.-long interval defined by the two ash-fall deposits contains ∼3.8 large-scale cycles (256 m), giving a cycle period of ∼395 k.y. This is close to the 413-k.y. period of the long-eccentricity cycle. The period for the short-term cycle (59 m) then equals ∼96 k.y.; this matches the main period of the short eccentricity.

The cyclicity interpretation of the power spectra in terms of time is shown in detail in Figures 7D and 7E. The subdued peaks at wavelengths of 146 m and 114 m are interpreted as first-order harmonics of the long-eccentricity cycle with cycle periods close to half of 413 k.y. and/or twice 96 k.y. (cf. Weedon, 1989). This harmonic relation is particularly clear when peaks are plotted versus cycle period (Fig. 7E), instead of frequency (Fig.7D). The subdued peaks with wavelengths of 74 m and 64 m and periods of 120 k.y. and 104 k.y. are within the short-eccentricity band (95–125 k.y.; De Boer and Smith, 1994).

The 145 individual cyclothems in the Dutch succession have a highly variable thickness, all on a sub-eccentricity scale. The complete succession is 1728 m long, and it equals ∼7 long-eccentricity or 31 short-eccentricity cycles, which amounts to ∼2.9 m.y. Based on a linear depth-thickness relation, cyclothem duration ranges between 5 and 32 k.y.; cyclothem periods are evenly distributed, and there is no dominant period. Some cyclothems have higher estimated durations up to ∼60 k.y. The average cyclothem measures 12.5 m, which equals 21 k.y.

The highly variable thickness of individual cyclothems suggests no specific control by precession or obliquity but points to cycle interference (De Boer, 1991; Sageman et al., 1997). The average duration of the cyclothems (21 k.y.) is close to the main precession period, which equaled 17 k.y. during the Pennsylvanian (Berger and Loutre, 1994), and the maximum duration for the bulk of the cyclothems (32 k.y.) is close to the main obliquity period (34 k.y.). This may indicate that the amplitude of precession-driven, sea-level fluctuations was sufficiently high to prevent obliquity and short eccentricity cycles from being recorded. The observation that 63 cycles in the Dutch section are represented by only 17 cycles in Kentucky indicates that only one out of almost four cyclothems (on average) was preserved in the Kentucky succession. Absence of part of the cycles is in line with the observation that regionally extensive erosion surfaces mark the base of cycles in the Appalachian Basin (Aitken and Flint, 1995). Chesnut (1997) noted that areas of low subsidence in the Appalachian Basin are characterized by 400 k.y. cycles; whereas in areas of higher subsidence in the same basin, 100 k.y. cycles have been preserved. If this trend of increased likelihood of preservation of shorter cycles is extrapolated to areas of even higher subsidence, it is expected that at some point sub-eccentricity cycles are preserved.

The comparable thickness of cyclothems in the two study areas indicates that cyclothem thickness was probably not controlled by subsidence rate, but instead by glacio-eustatic sea-level fluctuations. Because these were global and of higher magnitude (∼50–100 m) than typical subsidence rates (Soreghan and Dickinson, 1994; Soreghan and Giles, 1999; Rygel et al., 2008), different areas experienced the formation of accommodation space at relatively comparable rates during periods of long-term sea-level rise. This likely resulted in the formation of cyclothems of relatively comparable thickness in different areas on the globe. During subsequent periods of overall sea-level fall, non-deposition and/or erosion occurred. In areas of low subsidence these periods of non-deposition lasted longer than in areas where subsidence rates where high, resulting in less complete successions.

IMPLICATIONS FOR PENNSYLVANIAN CHRONOLOGY

Chronological interpretations of Pennsylvanian stratigraphy are based on an integration of conodont (Boardman and Heckel, 1989; Barrick et al., 2004; Heckel et al., 2007; Boardman et al., 2009) and ammonoid (“goniatites”) (Ramsbottom et al., 1978; Riley and Turner, 1995; Greb and Chesnut, 2009) zonations with radiometric ages (Burger et al., 1997; Davydov et al., 2004; Davydov et al., 2010; Schmitz and Davydov, 2012) and cyclostratigraphic interpretations (Heckel, 2008; Falcon-Lang et al., 2011). This has yielded a high-resolution time scale, which is of limited use however in the mainly terrestrial, and often barren, Middle and Upper Pennsylvanian successions of Europe and the Appalachian Basin. In those parts, stratigraphic subdivisions are based primarily on palynology and paleobotany (Ramsbottom et al., 1978; Cleal and Thomas, 1996; Peppers, 1996), and uncertainty exists as to how these successions tie into the global stratigraphic framework. This is reflected in many different correlations of the Westphalian substages to global Eastern European stages and North American regional stages, and in differences in interpreted duration of the individual substages (Heckel and Clayton, 2006; Greb et al., 2008; Heckel, 2008; Falcon-Lang et al., 2011; Van Hoof et al., 2013).

The large-scale cycle framework presented in this study was integrated with published radiometric ages for volcanic ash (“tonstein”) layers to refine the correlation of the Western European Westphalian substages, which are commonly used in the (terrestrial) Appalachian Basin as well (Greb et al. 2008). Because radiometric ages were determined using different methods (Ar/Ar and U/Pb), results cannot be easily compared and integrated (Davydov et al., 2004; Villeneuve, 2004). Most of the available radiometric ages are from the Ar/Ar isotope system (Fig. 8). They are consistent internally and between basins (Lyons et al., 1992, 2006). Recent radiometric ages are based on the U/Pb isotope system (Lyons et al., 2006; Eros et al., 2012). These are more accurate; but due to a different standard, they yield systematically older ages (Davydov et al., 2004).

To integrate ages from both isotope systems, Ar/Ar age determinations were shifted to match the U/Pb time scale, based on the Fire Clay reference bed from the Appalachian Basin. For the Fire Clay coal bed, both an Ar/Ar age (310.9 Ma) and an U/Pb age (314.6 Ma) are available (Lyons et al., 2006), which indicates that 3.7 m.y. must be added to Ar/Ar dates to fit the U/Pb scale. Note that the Fire Clay level (Late Duckmantian) is an interesting calibration point, because its U/Pb age of 314.6 Ma is very close to the U/Pb age of 314.4 Ma determined for the base Moscovian in Eastern Europe (Eros et al., 2012; Schmitz and Davydov, 2012) and could therefore serve as a tie point for North American, Western European, and Eastern European stratigraphy (Fig. 8).

It was found that the Duckmantian substage (Westphalian B) measures three long- and 13 short-eccentricity cycles, which amounts to 1.2 m.y. (Figs. 7 and 8). If the Fire Clay coal bed is taken as a reference level (314.6 Ma), the marine band at the base of the Duckmantian has an age of ca. 315.6 Ma, and the marine band at the base of the Bolsovian has an age of ca. 314.4 Ma. The intra-Duckmantian marine band (Maltby/Kendrick) then lines up at ca. 314.8 Ma (Fig. 7).

The Bolsovian substage (Westphalian C) in the Dutch succession represents at least three long-eccentricity cycles, which equals 1.2 m.y. However, the Dutch succession has been erosionally truncated, and the duration of the Bolsovian substage thus must be longer. Numerous Ar/Ar age determinations of volcanic ash-fall layers are available from the Bolsovian of Central Europe (Hess and Lippolt, 1986; Burger et al., 1997; Gradstein et al., 2004), and based on these, the total duration of the Bolsovian is estimated at ∼2.0 m.y. (Fig. 8). This places the base of the Asturian at ca. 312.4 Ma. Extrapolation of radiometric ages based on Burger et al. (1997) places the youngest Asturian rocks, which are truncated by a major unconformity, at ca. 310.8 Ma. Downward extrapolation from Stephanian radiometric age data gives a base Stephanian age of ca. 307.9 Ma and indicates that the duration of the Westphalian–Stephanian hiatus, in this case in the German Saar Basin, lasted some three million years (Burger et al., 1997). An Asturian-Stephanian hiatus characterizes the entire European Variscan foreland (Ziegler, 1990; Corfield et al., 1996; Schroot and De Haan, 2003), although commonly not shown in stratigraphic columns, and it relates to late Westphalian culmination of thrusting in the Variscan orogen (Ziegler, 1990). In the United States, a late Westphalian unconformity and associated hiatus are not recognized, or at least not thought to be as extensive as in Europe (Falcon-Lang et al., 2011).

Only a small part of the Langsettian (Westphalian A) substage is covered by the studied cores, and radiometric ages are not available below the Late Langsettian Upper Banner coal bed. Therefore an accurate estimate of the duration of the Langsettian is not possible. However, when the Langsettian thickness of 1200–1350 m (Drozdzewski, 1993) from the nearby Ruhr Basin in Germany (gray square in Fig. 2C) is extrapolated and constant subsidence is assumed, the Langsettian substage has an estimated duration of ∼2.0 m.y. This gives an estimated age for the base Langsettian of ca. 317.6 Ma (Fig. 8).

Based on the above discussion, the total duration of the Westphalian stage, at least the portion that has been preserved in Western Europe, lasted ∼6.8 m.y. The duration of the hiatus between the youngest Westphalian sediments (Asturian) and the overlying Stephanian section was ∼2.9 m.y., giving a maximum duration of Westphalian time of ∼9.7 m.y. This is considerably shorter than the earlier estimate of 11.5 m.y. by Menning et al. (2000). The shortening is primarily due to the newly available age data for the Late Langsettian Upper Banner coal bed. In the Geologic Time Scale 2004 of Gradstein et al. (2004), only 7.0 m.y. are assigned to the Westphalian stage, which is attributed to base Westphalian age estimation based on interpolation between Bolsovian (Westphalian C) and Namurian ash-fall ages and assuming similar sedimentation rates for the deeper marine Namurian and the paralic Westphalian depositional systems.

The Desmoinesian stage in the Midcontinent United States and the Upper Moscovian stage in Eastern Europe are commonly interpreted to partly overlap with the (Upper) Westphalian in Western Europe (Greb et al., 2008; Heckel, 2008; Falcon-Lang et al., 2011; Eros et al., 2012). However, the age determinations presented here place the entire Westphalian stage below the base Desmoinesian (309.4 Ma) of Heckel (2008).

TERRESTRIAL-CARBON (COAL) STORAGE

The Duckmantian and Bolsovian stages are the most prolific coal intervals in the Pennsylvanian and represent a large percentage of the terrestrial carbon buried during the late Paleozoic ice-house period (Berner, 2003). The accurately dated sedimentary record from the Netherlands allows a precise determination of the long-term burial of carbon (as coal) in one of the largest and fastest subsiding Pennsylvanian basins (Fig. 9). The cumulative-coal patterns are correlatable between wells and show low terrestrial-carbon accumulation rates (∼5 m coal/m.y.) before ca. 315.1 Ma (U/Pb) after which accumulation accelerated to ∼20 m coal/m.y.

The cumulative-coal pattern in Kentucky, although showing lower values, features a similar instant increase at 315.1 Ma. The patterns are almost identical when the Kentucky coal-abundance values are multiplied by the stratigraphic stretching factor of 2.8 (i.e., correcting for subsidence and/or preservation) that was required to align the marine bands in the high-subsidence Dutch succession and the low-subsidence Kentucky succession (Fig. 9). This suggests that overall coal accumulation was externally controlled by climate and/or sea level, and was comparable between the areas, the amounts of buried coal depending on subsidence, which also controlled the percentage of cycles deposited and preserved. From outcrops, it is known that coal thickness may show rapid lateral changes, but the results of this study show that this did not significantly affect the overall long-term coal-accumulation patterns.

DISCUSSION AND CONCLUSIONS

Since the beginning of the twentieth century, the cyclicity in Euramerica coal basins has been related to sea-level changes associated with late Paleozoic glaciations (e.g., Udden, 1912; Wanless and Shepard, 1936). Although it seems more than a coincidence that cyclothem formation was widespread and coeval with late Paleozoic glaciation, there was as yet no evidence that unambiguously demonstrates Milankovitch control. This is attributed here to the fact that previous analyses were based on cyclothem thickness, which is unlikely to reflect the sea-level fluctuations that shaped them, because—depending on the ratio between sea-level change, subsidence, and sediment accumulation—cyclothem boundaries may be formed almost anywhere during a sea-level cycle. Furthermore, previous studies were carried out in areas where the Pennsylvanian succession is relatively thin and incomplete, and cycle duration is typically estimated by dividing succession duration by the number of observed cycles. When many cyclothems are missing, cycle duration is grossly overestimated. This means that estimated cycle periods could be too long although still falling in the Milankovitch range (Algeo and Wilkinson, 1988).

It has been shown that Milankovitch signatures can be revealed by quantitatively analyzing variations in cyclothem composition through time, such as the subaerial-facies ratio and the coal percentage per cycle when done in high-subsidence areas where stratigraphic successions are most complete. In combination with radiometric ages, the data show a strong short- and long-eccentricity control, which is in agreement with current ideas (Eros et al., 2012; Waters and Condon, 2012). However, the data also show that the 100 k.y. estimate is too high for individual cyclothems and includes missed beats and/or eroded stratigraphy. Individual cyclothems have a sub-eccentricity duration, which is possibly the result of interference of precession, obliquity, and eccentricity signals. The interpreted dominance of precession cycles is remarkable when compared with δ18O records from the Quaternary; these records point to a dominance of obliquity (before 0.7 Ma) and short eccentricity (after 0.7 Ma). The influence of precession-controlled climate fluctuations is particularly strong at low latitudes (De Boer and Smith, 1994), and it is therefore not expected to play an important role during ice-house periods. However, a number of studies suggest that late Paleozoic ice sheets extended into low-latitude areas (Frakes et al., 1992; Poulsen et al., 2007; Soreghan et al., 2008), perhaps down to 30°–40° S during the Westphalian. At such low latitudes, precession-driven climate fluctuations may have been important in partially controlling or modulating the waxing and waning of ice caps and may have had sufficiently high amplitudes to break up obliquity cycles into smaller units. In a number of recent studies, however, the extent of ice sheets into relatively low-latitude areas is questioned; and especially during the late Westphalian and Stephanian, ice extent may have been limited (Isbell et al., 2003; Fielding et al., 2008a; Fielding et al., 2008b).

As an alternative, variations in sediment supply, either regional (allocyclic [cf. Cecil et al., 2014]) or local (autocyclic [e.g., delta-lobe switching; cf. Fielding, 1984]), may have been responsible for the splitting of sea-level cycles into shorter subcycles.

Integration of the cycle interpretations presented here with recent U/Pb ages from the Appalachian Basin allowed chronostratigraphic dating of the main marine bands at the base, middle, and top of the Duckmantian substage at 315.6, 314.8, and 314.4 Ma. This makes a robust chronostratigraphic framework for the Late Langsettian to Early Bolsovian interval that links the coal basins from Europe and North America. Based on Ar/Ar ages from Europe (recalculated to U/Pb), it is estimated that the entire Westphalian stage lasted ∼9.8 m.y. at most (including an ∼2.9 m.y. hiatus at its top) and is entirely older than the Midcontinent Upper Pennsylvanian key section of Heckel (2008).

The observation that the pattern of carbon (coal) storage through time can be correlated between the Dutch wells, and even between basins after correction for subsidence and missing cycles, is intriguing. It requires further study to establish whether the recorded patterns are indeed global rather than the result of local sedimentary conditions. If indeed correlatable over larger distances, coal patterns may be an interesting correlation tool and a possible proxy of paleoclimate and sea level.

The methodology applied here has the advantage that it is quantitative and therefore leaves less room for speculation when it comes to the interpretation of cycle periods. However, it requires thick, continuous stratigraphic successions in high-subsidence areas with a relatively complete sedimentary record. Based on a few good-quality sections, such as the early Langsettian “Joggins” sections of Nova Scotia, it may be possible to extend the Duckmantian–Bolsovian chronology presented here to the base of the Langsettian and into the Asturian, thus covering the entire Westphalian coal interval and giving a high-resolution time framework for a the major coal interval of the late Paleozoic glaciation.

We are thankful to Thomas Algeo and William DiMichele for their constructive reviews and to Poppe de Boer for his comments on an earlier version of the manuscript. Discussions with Hemmo Abels, Haefa Abdul Aziz, and Frits Hilgen (University of Utrecht) helped to improve the cyclicity analysis. Stephen Greb and Cortland Eble of the Kentucky Geological Survey (Lexington) are thanked for their assistance during our field trips and for sharing their ideas and expertise on the cyclothem successions of the Appalachian Basin.