Rainforest collapse triggered Carboniferous tetrapod diversifi cation in Euramerica Sarda Sahney1, Michael J. Benton1, and Howard J. Falcon-Lang2* 1Department of Earth Sciences, University of Bristol, Bristol BS8 1RJ, UK 2Department of Earth Sciences, Royal Holloway, University of London, Surrey TW20 0EX, UK ABSTRACT forest islands exerted a major impact on tetra- Abrupt collapse of the tropical rainforest biome (Coal Forests) drove rapid diversifi cation pod diversity, ecology, and the development of of Carboniferous tetrapods (amphibians and reptiles) in Euramerica. This fi nding is based on endemism. In doing so, we draw on the theory analysis of global and alpha diversity databases in a precise geologic context. From Visean to of island biogeography (MacArthur and Wilson, Moscovian time, both diversity measures steadily increased, but following rainforest collapse 1967), which was developed to explain pat- in earliest Kasimovian time (ca. 305 Ma), tetrapod extinction rate peaked, alpha diversity terns of diversifi cation in oceanic islands, but is imploded, and endemism developed for the fi rst time. Analysis of ecological diversity shows equally applicable to other kinds of islands, e.g., that rainforest collapse was also accompanied by acquisition of new feeding strategies (preda- rainforest refugia. tors, herbivores), consistent with tetrapod adaptation to the effects of habitat fragmentation and resource restriction. Effects on amphibians were particularly devastating, while amniotes METHODS: LATE PALEOZOIC (‘reptiles’) fared better, being ecologically adapted to the drier conditions that followed. Our TETRAPOD DATABASE results demonstrate, for the fi rst time, that Coal Forest fragmentation infl uenced profoundly In order to detect changes in tetrapod diver- the ecology and evolution of terrestrial fauna in tropical Euramerica, and illustrate the tight sity across the Moscovian-Kasimovian inter- coupling that existed between vegetation, climate, and trophic webs. val, we constructed two late Paleozoic tetra- pod databases, comprising records of global INTRODUCTION that show that sea level dropped to its one of and alpha diversity over nine global stages During the latter part of the Carboniferous its lowest levels in the entire Pennsylvanian, if (Visean, Serpukhovian, Bashkirian, Mosco- (318–299 Ma), Europe and North America not its lowest level (Heckel, 1991, 2008), pre- vian, Kasimovian, Gzhelian, Asselian, Sak- (Euramerica) were positioned over the equator, cisely coincident with the most abrupt phase marian, and Artinskian) ranging from 346 to and were covered, at times, by humid tropical of vegetation change (DiMichele et al., 2009). 270 Ma. We chose to restrict the analysis to this rainforest (DiMichele et al., 2007). This biome, An alternative hypothesis is that medium-term time span because the bracketing Tournaisian colloquially referred to as the Coal Forests, greenhouse warming drove aridifi cation, as and Kungurian stages were times of very low comprised a heterogeneous vegetation mosaic supported by far-fi eld records in Gondwana diversity, which have been interpreted as mass (Gastaldo et al., 2004) inhabited by a rich ter- (Fielding et al., 2008) and evaporites in high- extinctions or gaps in the record, i.e., Romer’s restrial fauna (Falcon-Lang et al., 2006). As cli- stand deposits in western Euramerica (Bishop gap and/or bottleneck (Ward et al., 2006) and mate aridifi ed through the later Paleozoic, these et al., 2010). However, regardless of what Olson’s gap and/or extinction, respectively rainforests collapsed, eventually being replaced caused aridifi cation, the consensus is that this (Sahney and Benton, 2008). by seasonally dry Permian biomes (Montañez et climate shift led to the fragmentation of the al., 2007). Collapse occurred through a series of Coal Forests into isolated rainforest islands Global Diversity Database step changes. First there was a gradual rise in the surrounded by xerophytic scrub (Falcon-Lang, Initially 67 families from 163 tetrapod sites frequency of opportunistic ferns in late Mosco- 2004; Falcon-Lang et al., 2009; Falcon-Lang worldwide were tabulated to create the global vian time (Pfefferkorn and Thomson, 1982). and DiMichele, 2010). diversity database. Analysis was run with all This was followed in the earliest Kasimovian At the time of peak levels of rainforest die- of the families and then was repeated after (cyclothem-calibrated age of 305.4 Ma; Heckel, back in the earliest Kasimovian, terrestrial removing 14 monotypic families, those repre- 2008) by a major, abrupt extinction of the domi- faunas had already become highly diversifi ed, sented by only a single species. The inclusion nant K-selected lycopsids and a switch to tree- composing sophisticated interconnected com- or exclusion of singletons made no difference fern dominance (DiMichele and Phillips, 1996). munities (Falcon-Lang et al., 2006). Detritivory to the results as they are randomly distributed In latest Kasimovian time, rainforests vanished was the most common primary feeding strategy through the time bins and the overall diver- (DiMichele et al., 2006). utilized by annelids, molluscs, and arthropods, sity patterns remained the same. Stratigraphic The nature and cause of late Moscovian- including the giant litter-splitting arthropleu- ranges were assigned to each family and the Kasimovian rainforest collapse have been the rids (Shear and Kukaloveck, 1990; Labandeira, associated dates were correlated with the subjects of intense investigation. In cratonic 2006). However, some insects had addition- Davydov et al. (2010) time scale. areas of North America (where the effects of ally evolved herbivorous and predatory forms Each family was also given an ecological tectonics can be excluded), an abrupt shift to (Labandeira and Sepkoski, 1993; Grimaldi and assignment based on size (snout-vent length; more arid climates has been linked to rainforest Engel, 2005). Terrestrial vertebrates (tetrapods), small: <0.15 m, medium: 0.15–1.50 m, large: collapse (DiMichele et al., 2009, 2010), though which included amphibians and basal amniotes >1.50 m) and diet (fi sh, insects, tetrapods, the exact causal mechanism remains uncer- (‘reptiles’), were mostly piscivores, refl ecting plants), resulting in 12 ecological niches. Diet tain. One hypothesis is that aridifi cation was their dominantly waterside habitats, but some was inferred from jaw and tooth structure, pat- triggered by a short-term but intense glacial forms also had evolved insectivory (Benton, terns of tooth wear, body size, and whether the phase. This is supported by earliest Kasimo- 2005; Coates et al., 2008). Here we analyze the animal was adapted for a predominantly aquatic vian paleosols in the Lost Branch cyclothem effects of rainforest collapse on tetrapod com- or terrestrial lifestyle (Benton, 1996). Occasion- munities. Specifi cally we test the hypothesis ally, direct evidence in the form of gut contents *E-mail: [email protected]. that population constriction into isolated rain- was available, e.g., conifer and pteridosperm © 2010 Geological Society of America. For permission to copy, contact Copyright Permissions, GSA, or [email protected]. GEOLOGY,Geology, December December 2010; 2010 v. 38; no. 12; p. 1079–1082; doi: 10.1130/G31182.1; 3 fi gures; 1 table. 1079 ovules in the Permian reptile, Protorosaurus (sometimes termed beta diversity, BD). Ende- (Munk and Sues, 1992). mism is calculated by dividing global diversity (Tt) by mean alpha diversity (T) (BD = Tt / T; Alpha Diversity Database Sepkoski, 1988). Community data, compiled in the alpha diversity database, were constructed as a subset RESULTS: TETRAPOD of the global database, containing the most com- DIVERSIFICATION PATTERNS plete tetrapod assemblages available. Individual Several patterns emerge from our analysis. assemblages were selected based on the occur- First, global diversity steadily rose through the rence of >100 partial skeletons at a given site study interval from 6 to 7 families in the Visean and, where possible, collector curves were used and Serpukhovian to 39 families in the Artin- to assess completeness of these assemblages. skian (Fig. 2A). However, while alpha diver- After fi ltering, the database contained 22 well- sity closely tracked global diversity until the sampled assemblages. late Moscovian (i.e., Nyrany and Linton alpha Although variably time averaged, we assumed sites), the two curves dramatically diverged that each assemblage was representative of a local across the Moscovian-Kasimovian boundary as community (sensu Begon et al., 2005). The num- alpha diversity collapsed from 20 families to 7 ber of families represented in each community families (Fig. 2A). was tabulated, based on published assignments Analysis of the rates of alpha and global (supplemental databases: global diversity data- diversifi cation helps explain this divergence. base—http://www.fossilrecord.net/fossilrecord/ Although the global diversifi cation rate slowed download.html; alpha diversity database—http:// across the Moscovian-Kasimovian boundary, the palaeo.gly.bris.ac.uk/Sahney/pub/index.html) as rate became strongly negative at the alpha (com- a proxy for alpha (community) diversity. Com- munity) level, the only time when either rate of munities were binned by stage, and assigned an diversifi cation became negative in the nine stages, average age (based on Davydov et al., 2010), to refl ecting the fact that communities shrank in construct an alpha diversity curve. size, i.e., an “alpha implosion”