Evidence for coal forest refugia in the seasonally dry Pennsylvanian tropical lowlands of the Illinois Basin, USA

Introduction

The Pennsylvanian Subperiod of the Carboniferous (323–299 Ma) was characterized by a series of glacial–interglacial cycles that exerted profound control on the distribution of vegetation at this time (Eros et al., 2012). In particular, these cyclic climatic changes resulted in the alternating dominance of wetland and dryland vegetation in the Pennsylvanian tropics (DiMichele, 2014). The wetland vegetation of this time period is represented in the fossil record by the classic Pennsylvanian Coal Forests, which were composed of arborescent lycopsids and, depending on the flooding regime, an understory of marattialean tree ferns, pteridosperms, sphenopsids and cordaitaleans (DiMichele & Phillips, 1994; DiMichele, 2014). In contrast, Pennsylvanian dryland vegetation is recorded by rare gymnosperm-dominated fossil assemblages, which contain drought-tolerant plants such as cordaitaleans, pteridosperms and walchian conifers (Falcon-Lang et al., 2009; Dolby, Falcon-Lang & Gibling, 2011). There is also evidence for the co-existence of wetland and dryland vegetation in at least regional proximity, reflected by fossil remains of dryland plants preserved alongside fossil wetland vegetation. These drier elements are thought to have been transported from upland (extrabasinal sensu Pfefferkorn, 1980) areas into wetland basins (e.g., Gastaldo, 1987; Lyons & Darrah, 1989; Falcon-Lang & Bashforth, 2004; Gastaldo & Degges, 2007).

A good example of Pennsylvanian dryland vegetation is the Moscovian age macroflora found at Cottage Grove Mine, Illinois (Falcon-Lang et al., 2009). This flora was discovered in fine siltstone layers closely associated with conglomeratic facies within a shallow channel, laterally equivalent to a calcic vertisol (indicating a seasonally dry climate) just below the Baker Coal in southeastern Illinois, USA (37°46′N, 88°25′W) (Feldman et al., 2005; Falcon-Lang et al., 2009). The Cottage Grove macroflora is composed mostly of cordaitaleans together with pteridosperms, walchian conifers, ferns and sphenopsids, and lacks classic Coal Forest plants such as lycopsids (Falcon-Lang et al., 2009). The small size of the channel (∼250 m wide) and locally derived channel fill indicates that the drainage area was relatively small (Feldman et al., 2005). The plant macrofossils (leaves and other vegetative and reproductive structures) at Cottage Grove are associated with locally derived conglomerate and this, together with information on channel morphology, is interpreted as evidence that the macrofossils represent local vegetation growing on interfluves, close to the depositional environment (Falcon-Lang et al., 2009).

There is currently no palynological data from the same stratigraphic horizons as the plant macrofossils that are preserved at Cottage Grove (Falcon-Lang et al., 2009). Reconstructions of the standing vegetation at this locality therefore lack the complementary data that can be provided by sporomorphs (prepollen, pollen, megaspores and spores) (e.g., Chaloner, 1968; Gastaldo et al., 1998; Jackson & Booth, 2007; Mander, Kürschner & McElwain, 2010). To address this deficit, we have undertaken a palynological investigation of the macrofossil-bearing shale found within the conglomeratic channel facies at Cottage Grove. This shale, and the sampled interval, is important because it was not deposited in association with widespread wetland environments typical of peat forming portions of glacial-interglacial cycles. Our results highlight considerable differences between the macrofossil and sporomorph records at Cottage Grove. Notably, our study reveals a population of 17 species of lycopsids from the orders Isoetales, Lepidodendrales and Selaginallales that flourished at Cottage Grove, but is entirely absent from the macrofossil record at this locality. Consequently, suggestions that Pennsylvanian dryland vegetation at Cottage Grove was dominated by xerophytic plants and devoid of wetland taxa such as lycopsids (Falcon-Lang et al., 2009) are incompatible with our palynological data. We interpret the discrepancy between the macrofossil and sporomorph records at Cottage Grove as evidence that the Pennsylvanian dryland interglacial landscape in this region included fragmented populations of wetland plants living in refugia.

Materials and Methods

Plant macrofossils at Cottage Grove are preserved in fine-grained siltstones in close association with conglomeratic layers (Falcon-Lang et al., 2009). We have analyzed the palynological composition of two samples from sediments containing a conifer macrofossil that occurred in a layer sandwiched between channel-bottom siltstones and conglomeratic deposits (lithologically unit 6a at the Cottage Grove locality; see Fig. 2A of Falcon-Lang et al., 2009). We have taken a palynological sample from the sediments (Sample 1) in which the conifer macrofossil (USNM 536629; see Fig. 3 of Falcon-Lang et al., 2009) was embedded, and a palynological sample of the counterpart surface (Sample 2). This means that the deposition of the two samples was temporally separated by less than the amount of time it would have taken for the deposited conifer remains to decay. Sample 2 originated from a slightly darker area on the counterpart surface of the same conglomeratic unit.

The sample was bulk macerated using hydrofluoric and hydrochloric acid, and the remaining organic residue was processed at the Laboratory of Palynology and Palaeobotany, Utrecht University, according to palynological techniques (heavy liquid separation and sieving over 15 µm mesh). The samples were not sieved over 250 µm mesh. This will be addressed later. The organic residue was mounted in glycerin jelly. The two samples were screened for identifiable sporomorphs, and a count of 433 (Sample 1) and 300 (Sample 2) sporomorphs was performed at the species level. Additional slides from the two samples were then screened for rare species. Two additional rare taxa were recorded in this process. Taxonomic descriptions by Smith & Butterworth (1967), Ravn (1979) and Ravn (1986) were used for identification. A list of sporomorph taxa, their botanical affinities, and their relative abundances in each of the two samples were compiled (Table 1, Table S1).

Results

The sporomorph assemblage of Sample 1 is dominated by characteristic Middle Pennsylvanian wetland taxa (Table 1). Spores of arborescent lycopsids (e.g., Cadiospora, Lycospora spp., Granasporites) make up the majority of the assemblage (60.8%), with fern spores (16.9%) and cordaitalean prepollen (Florinites spp., 15.5%) as the other major components (Fig. 1). Notable occurrences include Lycospora granulata, which was produced by the highly specialized wetland arborescent lycopsid Lepidophloios hallii (DiMichele, 2014), and represents 15.2% of the total sporomorph assemblage. Sub-arborescent lycopsids are represented by low numbers of Endosporites, Radiizonates and Cristatisporites. Spores that have been found in situ herbaceous lycopods (Cirratriradites spp.) are rare elements. Sample 1 also contains the spores of a variety of ferns, including marattialean tree ferns (Cyclogranisporites, Thymospora, Punctatosporites, and Latosporites) as well as spores of sub-arborescent Calamites (Calamospora) and smaller-sized Sphenophyllales. Cordaitaleans are represented by four Florinites species, of which F. mediapudens is the most abundant (10.6%). Prepollen of other seed plants, such as medullosan prepollen (Schopfipollenites), walchian conifer prepollen (Potoniesporites), or other pseudosaccates or bisaccates are absent from among the ∼1,500 palynomorphs that have been scanned. The diversity of this sample is 39, and its evenness is 0.76.

In Sample 2, the cordaitalean prepollen Florinites dominates the sporomorph assemblage (50.3%), while spores of ferns (28.3%) and arborescent lycopsids (16.0%) are also abundant (Fig. 1). In this sample, Granasporites medius is the only abundant spore produced by the Lepidodendrales. Sigillarian lycopsid spores (Crassispora kosankei) are common, and spores of herbaceous lycopsids (Cirratriradites) are rare in this sample. Marattialean tree ferns are represented by Cyclogranisporites (5.3%), and sub-arborescent Calamites is represented by Calamospora. Spores of other ferns and the smaller-sized Sphenophyllales are present, but in low numbers. Prepollen produced by other seed plants, conifers and seed ferns, is not present in this sample. The diversity of this sample is 32, and its evenness is 0.69.

Comparison of samples

The sporomorph assemblage of Samples 1 and 2 are composed of the same plant groups, but the relative abundance of these groups in the two samples is substantially different (Fig. 2). Although sphenopsid spores are a minor component of sporomorph assemblages in Samples 1 and 2, cordaitaleans and ferns are much more abundant in Sample 2 than in Sample 1, and lycopsids are considerably more abundant in Sample 1 than in Sample 2 (Fig. 2). There are also major differences in the species-level composition of the two samples. For example, of the 54 sporomorph species that were recorded in the two samples, just 17 are present in both (Table 1), and a Sorensen’s index comparison of Sample 1 and Sample 2 returns a value of 0.48 (Table 2) (SI = 20C/[A + B], where C is the number of species in common between two samples, and A and B are the total number of species in each of the two samples (McElwain et al., 2007)). These differences are surprising given the stratigraphic proximity of Sample 1 and Sample 2. Sample 2, however, originated from a slightly darker area on the counterpart surface of the same conglomeratic unit, and it is possible that this reflects subtle differences in the taphonomic conditions of the two samples. Factors that could create compositional differences between the two samples include the hydrodynamic regime (Havinga, 1967), the degree of oxidation or microbial activity, and the action of wet and dry cycles (Campbell & Campbell, 1994; see Mander et al., 2012 for a review).

Table 2:

Abundance major plant groups in macrofossil and sporomorph assemblages.

Relative abundance of major plant groups in the macrofossil and sporomorph records at the Cottage Grove plant locality, Cottage Grove Mine, southeast Illinois. See Table 1 for count data.

Plant group	Macrofossils (%)	Sample One (%)	Sample Two (%)
Cordaites	68.3	15.5	50.3
Lycopsids	0.0	62.6	16.3
Sphenopsids	2.4	4.4	5.0
Ferns	9.8	16.9	28.3
Conifers	7.3	0.0	0.0
Pteridosperms	12.2	0.0	0.0

DOI: 10.7717/peerj.630/table-2

Discussion

This comparison of the macrofossil and sporomorph records at Cottage Grove builds upon previous comparisons of these two fossil groups in the Carboniferous (e.g., Mahaffy, 1988; Willard, 1993; DiMichele & Phillips, 1994), and emphasizes that they provide very different pictures of the standing vegetation at a given locality (e.g., Chaloner, 1968; Gastaldo et al., 1998; Jackson & Booth, 2007; Mander, Kürschner & McElwain, 2010). Especially the absence of prepollen of medulosan pteridosperms (e.g., Schopfipollenites and Monoletes) and walchian conifers (Potonieisporites) in Cottage Grove’s palynological samples is striking. It has been suggested that certain differences between the pictures provided by each fossil group may be explained by palynological processing techniques. For example, large (>200 µm) prepollen referable to medullosan pteridosperms can be inadvertently removed from palynological preparations by sieving over coarse mesh (DiMichele & Phillips, 1994). However, our samples were only sieved over 15 µm mesh, not 200–250 µm mesh as is standard in some palynological processing protocols. Therefore, in contrast to previous suggestions (DiMichele & Phillips, 1994), sieving cannot explain why these plants are present in the macrofossil record at Cottage Grove (Falcon-Lang et al., 2009), but entirely absent from our palynological samples from the same sediments (Fig. 3; Tables 2 and 1). It is possible that our samples, which are derived from very thin siltstone horizons, represent very brief intervals of geological time in which there was no conifer pollen being produced by plants in the source vegetation.

Other differences between the macrofossil and sporomorph records of vegetation at Cottage Grove could be explained by the different ways in which macrofossils and sporomorphs are representative of the standing vegetation, in terms of composition, dominance and spatial scale. In general, macrofossil assemblages are typically weighted towards plants of large stature that produce a considerable number of potential fossils that may disperse widely (e.g., Spicer, 1989; Greenwood, 1991; Gastaldo, 2001). In addition, parautochthonous macrofossil assemblages will tend to be strongly biased toward those plants living in close proximity to the environment of deposition (Scheihing & Pfefferkorn, 1984; Burnham, Wing & Parker, 1992). In contrast, sporomorph assemblages are weighted towards taxa with high sporomorph productivity and/or taxa that produce sporomorphs that are deposited slowly from the atmosphere (Prentice, 1985). Sporomorphs also can be carried longer distances from source areas by water than is typical for foliage, and certain depositional settings, such as large lakes, therefore may have a significant extra-local elements not represented among the macrofossils (Farley, 1990).

In addition to the absence of conifer and pteridosperm prepollen, there is a distinct mismatch between cordaitalean abundance in the Cottage Grove macro and microfloras. The dominance of Cordaites leaves in the macrofossil assemblages at Cottage Grove (Falcon-Lang et al., 2009; Fig. 3; Table 2) may be partially a taphonomic bias, reflecting their robust, “leathery” construction (Stewart & Rothwell, 1993, p. 400). Their morphology made them differentially resistant to destruction, particularly if carried in relatively coarse, bed-load sediment. There are suggestions of a counter bias with regard to cordaitalean pollen. Comparison of macrofossil and sporomorph assemblages in Westphalian Coals indicated that sporomorphs produced by cordaitaleans were under-represented relative to macrofossils of the same genus (DiMichele & Phillips, 1994), an observation consistent with our data (Fig. 3; Table 2).

The opposite of the walchian conifer and medullosan pteridosperm pattern is that of the lycopsids. The microfossil-macrofossil mismatch in terms of lycopsid abundance in Cottage Grove Sample 1 (Fig. 3) only partially reflects the high abundance of three Lycospora species in sporomoph sample (Table 1). Over-representation of Lycospora relative to macrofossil estimates of parent-plant abundances also matches comparisons from Westphalian coals (DiMichele & Phillips, 1994). The relative over-representation of lycopsids in Sample 2 from Cottage Grove (Fig. 3; Table 2), however, is mainly due to the high abundance of Granasporites in this sample (Table 1). This is unexpected because Granasporites is thought to be under-represented in sporomorph assemblages due to low spore productivity by the parent plants, Diaphorodendron and Synchysidendron (DiMichele & Phillips, 1994). This may indicate that some results from macrofossil–sporomorph comparisons in peats and coal balls (e.g., DiMichele & Phillips, 1994) cannot be generalized, but it is more likely that there were cryptic populations of Granisporites producing plants nearby on the landscape.

Supplemental Information