Pennsylvania:! 'fossil forests' in growth position (T° assemblages): origin, taphonomic bias and palaeoecological insights

Fossil forests, buried in growth position in a geological instant (T° assemblages) are far more abundant in Pennsylvanian successions than in any other part of the geological record. In this review paper, we evaluate the fundamental controls on the origin of these phenomena, investigate the taphonomic biases that influence their composition, and summarize their palaeoecological significance. Following earlier workers, we highlight that high rates of burial and accommodation are essential for the formation and preservation of T° assemblages. Contexts especially favourable for their origin include ashfalls proximal to volcanic centres, coastal plains drowned by relative sea-level rise, and fluvial environments such as channel bars, crevasse splays, and distributary lobes. Long-term preservation requires high rates of subsidence. Consequently, the vast majority of Palaeozoic T° assemblages are confined to wetland settings at, or close to, sea level, whereas drylands are poorly represented and uplands rarely sampled, if ever. However, this is not the only major bias in the fossil record; taphonomic processes selectively preserve plants dependent on their anatomy and stature, and on groundwater chemistry. Thus, although T° assemblages offer unrivalled insights into the nature of ancient forests (whole-plant reconstructions, tree density, canopy height, productivity, plant hydraulics, cohort dynamics, spatial heterogeneity, ecological gradients, tree-sediment interactions, and animal-plant interactions, to name but a few), it is naive to believe they provide 'photographic snapshots' of palaeoecosystems. None the less, careful taphonomic analysis of T° assemblages offers the potential for a nuanced understanding of these evocative phenomena, and much remains to be learned from these important palaeoecological resources. Single fossil plants to entire forested landscapes (DiMichele et al. 2007) are commonly buried in place in the post-Silurian sedimentary record. For over two centuries, these phenomena have captured the imagination of natural scientists and the general public alike. They are the closest thing in the botanical world to the human landscapes buried at Pompeii and elsewhere (Deem 2005), which preserve a record of human 'ecology', and so fascinate us. The famous Carboniferous lycopsid tree stumps enshrined at Victoria Park in Glasgow, Scotland (Fig. 1; MacGregor & Walton 1948; Gastaldo 1986a), for example, have reminded us that there were trees living in the past with which we today have no familiarity and of which we have only limited understanding. Such remains are at the same time familiar and bizarre, the latter ever more so the deeper in time one looks. Throughout this paper we will use the term 'tree' in a relatively colloquial sense, to include large-bodied, upright plants with a main trunk or trunks and most often a crown of some form. Most modern trees are woody, and most definitions of the term 'tree' include this as a diagnostic feature. However, trees of the Late Palaeozoic were not so uniformly woody, although many were giants, many metres in height, supported by specialized bark (Boyce et al. 2010), mantles of supporting roots (Ehret & Phillips 1977), or even by leaning on one another, while living in tangled thickets (Wnuk & Pfefferkorn 1984). The situation becomes stranger, the further back in time we retreat. The recent discoveries of weird 'trees' in Devonian strata, including arborescent fungi (Boyce et al. 2007) and cladoxylopsids (Stein et al. 2007; Meyer-Berthaud et al. 2010), well illustrates this point. For the study of palaeoecology, trees preserved in growth position provide a glimpse of the spatial structure of ancient ecosystems, something not readily accessible from other kinds of fossil preservation. Nothing gets us closer to the original landscape and its plant cover or reveals so much about plant size, habit, or even ontogeny (e.g. Libert in et al. 2009). In this paper, we employ the term 'T° assemblages' (after Gastaldo et al. 1995, 2004a, 2006; Johnson 2007; DiMichele & Gastaldo 2008) to describe fossil forests preserved in essentially the same spatial conformation as in life. This includes assemblages buried in place; that is, the remains of upright trees rooted in their original soils (autochthonous) and assemblages where all or some of the vegetation has been slightly displaced (e.g. undisturbed leaf litter layers; Burnham et al. 1992) but still preserving most of the original vegetational structure (parautochthonous; Gastaldo 19866; Gastaldo et al. 1987; Behrensmeyer et al. 2000; MartinClosas & Gomez 2004). Most parautochthonous assemblages have been transported to some small degree from their original spatial position (Bateman 1991) and, therefore, do not preserve enough original spatial information to permit the drawing of unambiguous inferences about the interrelationships of different plants or groups of plants. Most root-containing palaeosols are also excluded, mainly because plant remains preserved within them may be highly time-averaged, reflecting the intermixing of root systems from different generations of plants (as may be demonstrated from cross-cutting relationships in some coal balls; DiMichele & Phillips 1994). It is an interesting but little discussed fact that T° assemblages have been reported far more often in Pennsylvanian successions (318-299 Ma) than in any other interval in the geological record. It is unclear whether this is because they are in fact much more abundant or simply more readily observed, a consequence of the