The plant fossil record reflects just two great extinction events

Introduction

The concept of mass extinctions was originally developed to describe the disappearance of large numbers of organisms over relatively short geological spans of time (Raup and Sepkoski, 1982; Sepkoski, 2001; Foote and Miller, 2007). Such extinctions are generally considered to have been a consequence of catastrophic events or major environmental changes that occurred too rapidly for organisms to adapt (Raup, 1994; Jablonski, 2005; McElwain and Punyasena, 2007). A large number of extinction events have been registered for the Phanerozoic Eon, but they are only assigned the status of mass events when they significantly exceed background extinction rates (i.e. they cross the upper boundary of the confidence interval of extinction pattern). Raup and Sepkoski (1982) used a simple form of time-series analysis at the rank of family to distinguish between background extinction levels and mass extinctions in marine faunas, and identified what are today known as the five major extinctions in Earth's history. These were interpreted as having had a global impact and, although they were defined only using marine fossil data, it was assumed that they represented deep perturbations affecting all biota. These events undoubtedly caused ecological disruption to vegetation (Wing, 2004; McElwain and Punyasena, 2007), but it has been argued that plant clades may have been more resilient to such events than animal clades (Traverse, 1988). Previous analyses of the plant fossil record appeared to corroborate this view (Cascales-Miñana and Cleal, 2012), but these analyses were based on a temporal resolution of epoch, which is too coarse for a comprehensive assessment of mass extinctions. Here we present a new analysis of the fossil record of vascular plants (tracheophytes) at the resolution of age/stage, following the approach used by Raup and Sepkoski with the marine fossil record (Raup and Sepkoski, 1982), to identify mass extinctions among plants independently of the perspective provided by the marine realm.

Data

A new comprehensive family-level dataset was compiled to provide a means of investigating origination–extinction processes and taxonomic diversity changes in vascular plants. Data compilation was conducted at the age-level of temporal resolution (i.e. using time intervals equivalent to chronostratigraphical stages). Absolute ages and interval lengths were extracted from Cohen et al. (2013). Information about early flora, (e.g. rhynophytes, zosterophylls, trimerophytes) was taken from Boureau et al. (1967), Cleal (1993a), Li (1995), Gensel and Edwards (2001), Raymond et al. (2006), Taylor et al. (2009), Hao and Xue (2013) and Cascales-Miñana and Meyer-Berthaud (2013). Palaeofloristic data about pteridophytes were extracted firstly from Cleal (1993a) and then modified by following Li (1995), Collinson (1996), Cleal and Thomas (1999a) and Taylor et al. (2009). This group was subsequently adjusted from the primary literature (Rössler and Galtier, 2002; Hilton et al., 2004; Tomescu et al., 2006; Kustatscher et al., 2007, in press; Shi et al., 2010; Decombeix et al., 2011; Galtier et al., 2011; Cariglino et al., 2012; Cariglino, 2013; Wang et al., 2014). Gymnosperm data were extracted from Cleal (1993b) and Anderson et al. (2007) with modifications from Cleal and Shute (2003), Uhl (2004), Kustatscher et al. (2007), Wang et al. (2007), Wang and Pfefferkorn (2010) and Pott (in press). For angiosperms, we firstly consulted Collinson et al. (1993), which was then updated following Friis et al. (2011). The incorporation of data for Cenozoic angiosperms at stage-level is tentative but, as these data refer only to first appearances, they do not affect extinction rates. Palynological data were not used. The resulting compilation encompasses 387 plant families along 76 time units. Raw data are available in Data S1.

Methods

Data analysis was conducted at family-level based on presence/absence data. Ranges were treated as continuous between the first and last appearances (Boltovskoy, 1988). To determine variations in family extinction levels through time in the plant fossil record, we calculated, for each time unit, the number of extinctions per Ma (see Table 1 in Cascales-Miñana et al., 2013). Omitting time units in which no taxa became extinct, we subjected these extinction rates to a regression analysis and used the 99 per cent confidence interval around the resulting regression line to represent the variation in background extinction rates (Raup and Sepkoski, 1982; Hammer and Harper, 2006). Excursions from this background rate were interpreted as significant extinction points (Raup and Sepkoski, 1982). The impact on the biota of the significant extinction points was measured by comparing the taxonomic severity (extinctions per total diversity per stage) (McGhee et al., 2012) with an estimation of mean standing diversity (total taxa discounting half of the originations and extinctions per stage) (Foote, 2000). Plant groups that are rarely preserved were not included in the computation.

To determine whether interval duration had any impact on the perception of such events, we also tested on the entire dataset the correlation between the duration of the time units and the total number of extinctions. This analysis was performed using R Software (R Developmental Core Team, 2013). Pearson (r), Spearman (r_s) and Kendall (τ) coefficients were used (e.g. Fröbish, 2013). The correlation was investigated both on raw data and on data transformed using generalized differencing that attempts to avoid false positives resulting from overall trends in the two datasets (see http://www.graemetlloyd.com/methgd.html for implementation). The significance level was fixed at 1%. A polycohort analysis was also included to illustrate how families progressively became extinct. This allowed us to test for the presence and effect of any extrinsic stress on diversity (Cleal et al., 2012). See Cascales-Miñana and Cleal (2012, and references therein) for details.

Results

The extinction pattern of tracheophytes between the Silurian and present day is shown in Fig. 1A, together with the resulting regression line and its 99 per cent confidence interval. This reveals that the normal extinction rate for plants is less than 1.0 family per Ma. Figure 1A highlights just four outliers with extinction levels significantly above the upper limit of 99 per cent confidence for the regression. The maximum excursion corresponds to six extinctions per Ma registered in the Changhsingian Stage (Lopingian). The other three high extinction points appear at the Kasimovian (Upper Pennsylvanian), Asselian (Cisuralian) and Roadian (Guadalupian) stages, with values of 1.8, 3.2 and 2.0 families per Ma respectively. Figure 1B shows the curve of standing diversity for plant families. This reveals a continuous decline in plant diversity through the Upper Pennsylvanian Series and the Permian System, a stratigraphical interval delimited by the outliers shown in Fig. 1A.

No significant correlation was found between interval length and the observed extinctions, in either the raw data (r = 0.01; P > 0.01; r_s = 0.15; P > 0.01; τ = 0.11; P > 0.01) or the generalized-differenced (detrended) data (r = −0.05; P > 0.01; r_s = 0.09; P > 0.01; τ = 0.03; P > 0.01). Diversity data revealed, however, that the strongest deflections of the survivorship curves occurred simultaneously with the extinction events described above (Figure S1). Polycohort analysis also revealed important deflections at the Devonian-Carboniferous and Triassic-Jurassic transitions, although these losses of diversity were within background extinction levels according to Fig. 1A.

Discussion

Our results provide evidence compatible with there having been two mass extinctions in tracheophytes, but in neither case were they discrete events. The first of them is detected at the Carboniferous-Permian transition, which is interpreted as a consequence of the collapse of the tropical wetlands in Euramerica (Cleal et al., 2010, 2011; Sahney et al., 2010), which hosted the taxonomically most diverse known vegetation of Pennsylvanian age (Cleal, 1991). The two-phase nature of this taxonomic extinction probably reflects the complexity of the environmental changes that were taking place during the late Moscovian-early Sakmarian time interval (c. 10 Ma). Although a major decline of the wetlands took place during late Moscovian times, there were refugia for the vegetation in a number of European intramontane basins (Doubinger et al., 1995), and lowland paralic basins in Iberia (Wagner and Alvarez-Vazquez, 2010), and in North America (DiMichele et al., 1996; Willard et al., 2007). However, these refugia appear to have been unable to maintain the diversity of plants that was seen during Moscovian times, and Kasimovian times saw the disappearance of a number of the families that had helped characterize the Moscovian tropical wetlands. By Asselian times, even these refugia had disappeared from Euramerica and, although similar tropical wetlands were now covering large areas of China (Hilton and Cleal, 2007), many of the characteristic Carboniferous wetland families disappeared at this time (e.g. Flemingitaceae, Diaphorodendraceae, Tedeleaceae, Urnatopteridaceae, Alethopteridaceae, Cyclopteridaceae, Neurodontopteridaceae). Altogether, half of the plant families disappear from the fossil record during either the Kasimovian or the Asselian stages.

The second mass extinction corresponds to the end-Permian event identified in the marine faunal record, but here again the process was spread over a significant time interval, in this case c. 20 Ma. The process started in Roadian times in tropical wetland vegetation from South China (e.g. Sigillariostrobaceae, Botryopteridaceae, Asterothecaceae, Tingiostachyaceae, Callistophytaceae) (Rees, 2002). Interestingly, however, there were also changes at this time in the southern middle to high latitude flora with, for instance, extinctions of Arberiaceae (Retallack et al., 2006). It has been suggested that in North China this was a relatively protracted event (Stevens et al., 2011) (although this still would not be resolved by our stage-level analysis) and may have been triggered by multiple environmental changes. However, the fact that it also appears to have coincided with changes in higher latitude vegetation suggests that more global, perhaps climatic, influences were in play. It is possible that we are seeing here a scenario comparable to that envisaged in late Pennsylvanian times where changes in tropical wetland vegetation in Euramerica triggered climatic changes that had global effects (Cleal and Thomas, 1999b, 2005; Cleal et al., 2011).

Results reveal the last major reduction in vascular plant diversity was at the Permian-Triassic boundary, with a loss of about 55% of plant families. The true magnitude of such events can be difficult to establish when looking at restricted palaeogeographical areas (Ruban and van Loon, 2008), but in this case it can be identified by family extinctions in the vegetation of the tropical wetlands (e.g. Lepidocarpaceae, Gigantopteridaceae) (Wang, 1989, 2010; Stevens et al., 2011), southern latitudes (e.g. Ottokariaceae) (Retallack et al., 2006) and northern latitudes (e.g. Rufloriaceae, Tchernoviaceae, Cardiolepidiaceae) (Goman′kov, 2005). It has been suggested there was also a marked reduction in species and genus diversity among tracheophytes based on a stage (or two stage) resolution analysis (Xiong and Wang, 2011). In contrast, other studies on the Chinese flora at a finer resolution suggest a more complex pattern (Wang, 2010; Stevens et al., 2011).

The extinction of marine families identified across the Frasnian-Famennian boundary (Raup and Sepkoski, 1982) coincided with the disappearance of a number of early vascular plant families. However, this elevated value is still within the confidence limits for background extinction rates for tracheophytes. These families may have disappeared because they were being out-competed by newly evolved taxa with innovative growth (Galtier and Meyer-Berthaud, 2006) and reproductive strategies (Linkies et al., 2010) rather than as a result of any significant disruption to habitats.

Another biotic crisis identified among marine taxa and terrestrial vertebrates at the end of Serpukhovian times (the ‘mid-Carboniferous boundary’) (Saunders and Ramsbottom, 1986; Weems, 1992) has been compared to the ‘Big Five’ extinctions (Bambach et al., 2004). It has been interpreted as being a major ecological crisis (McGhee et al., 2004, 2012), probably linked with global cooling and the expansion of the Gondwana ice-sheet (Fielding et al., 2008). No equivalent crisis can be discerned in the distribution of plant families and, if anything, this was a time of diversification in vegetation (Fig. 1B). This probably reflects the expansion of lowland habitats as the expansion of polar ice caused sea-levels to fall; such habitats being favourable to the relatively primitive middle Carboniferous vegetation.

Finally, the two post-Palaeozoic mass extinctions recognized in the marine fossil record, at the Triassic-Jurassic and Cretaceous-Palaeogene boundaries, also find little significant expression in our data. A slight increase in extinctions among plant families has been reported at the Triassic–Jurassic boundary (Cascales-Miñana and Cleal, 2012), but the relatively long duration of the Rhaetian Age means that the extinction rate per million years was not significantly higher than background rates. At the Triassic-Jurassic boundary, polycohort analysis has also shown a reduction in diversity, probably driven by the effect of the large number of short-lived taxa in the Carnian (Anderson et al., 2007). A similar lack of reduction in insect diversity has also been noted at this time (Labandeira and Sepkoski, 1993). In contrast, a loss of several tetrapod vertebrate families has been reported through the Upper Triassic Series (Benton, 1987, 1989), although it has been suggested that the extinction rates at the end of the Triassic Period were not anomalously high (Weems, 1992).

The iconic end-Cretaceous biotic crisis had a significant effect on marine and terrestrial faunas (Raup and Sepkoski, 1982; Sepkoski, 1982; Alroy, 2008; Schulte et al., 2010; Longrich et al., 2011, 2012; Mitchell et al., 2013), and caused ecological disruption and localized loss of species diversity in vegetation (Wilf and Johnson, 2004; McElwain and Punyasena, 2007; Nichols and Johnson, 2008; Mizukami et al., 2012). However, all tracheophyte families that have been recognized in latest Cretaceous fossil flora are extant today.

Conclusions

We have shown how the plant fossil record clearly reveals a different pattern of major taxonomic extinctions compared with that of faunas, especially marine organisms. Only two major extinctions can be recognized in plant families, and only one of these coincides with a major extinction in the marine realms. There has, moreover, been no major post-Palaeozoic extinction event among vascular plants. Plant taxa have been far more resilient than animal taxa to many types of major ecological disturbances such as pollution caused by large igneous province eruptions or bollide impact. On the other hand, plant taxa appear to have been impacted by events such as large-scale landscape disturbance (Cleal et al., 2011) that had a relatively limited effect on marine habitats. An important factor here may also be time-scale: if the duration of the ecological disruption did not exceed that of the viability of the disseminules of the plants (seeds, spores), those plant taxa had the potential to recover (Traverse, 1988), whereas the same relatively short-lived disruption could reduce the population size of animal taxa below what was viable for their long-term survival. This raises the question, can we really refer to an event as a mass extinction if it only has a taxonomic effect on a relatively limited set of organisms, albeit a set that tends to dominate the observed fossil record?