Biological principles suggest that recorded species diversity should increase as sampled area and habitat diversity increase. However, the interactions between Earth system drivers of these physical quantities, their representation in the preserved rock record, and fossil species counts are debated. We tracked marine bedrock area, lithofacies diversity, and echinoid species diversity through an 80 m.y. sea-level cycle, as recorded in the Cretaceous of the United Kingdom, to test these relationships. Relative completeness of the rock record rises during transgression but drops rapidly at maximum flooding and continues to fall during highstand. Lithofacies diversity peaks during transgression but drops sharply as maximum flooding is reached, whereas bedrock area peaks at maximum flooding then declines progressively through the highstand interval as a result of post-Cretaceous erosion. Both bedrock area and number of lithofacies units correlate with species diversity, but species diversity is predicted best by the two combined. There are no simple relationships between bedrock area and lithofacies diversity or the original area of marine sediments and range of habitats that once existed; these variables are strongly shaped by stratigraphic architecture. While the geological record left by a major sea-level cycle accurately captures the biological evolution of local communities, it is incomplete and systematically biased in its coverage of environments and communities. Erosional truncation of sediment packages and species ranges in the later stages of sea-level cycles explains the reported asymmetric relationship between macroevolution and macrostratigraphy.


Over the past 10 years, the link between sampled fossil diversity and the structure of the geological record has come under the spotlight. Whether using formation numbers, bedrock area, numbers of sedimentary packages, or counts of sections, a remarkably strong positive correlation emerges between rock availability at outcrop and the recorded diversity of fossils (see McGowan and Smith, 2011, for a compilation of papers). Although this relationship was first quantified by Raup (1976), controversy remains concerning its cause. The three potential explanations are not mutually exclusive: (1) a sampling artifact acting through a species/area effect (Raup, 1976; Smith, 2001); (2) diversity and rock area responding independently to the same common driver (major sea-level cycles; the common-cause hypothesis of Peters, 2005); (3) redundancy, i.e., a reversed causality hypothesis in which species diversity drives the number of fossil localities (Benton et al., 2011). Because these potential drivers are probably acting in concert, disentangling biological signal from sampling bias in the fossil record has proven to be very difficult. Large-scale sea-level cycles have been shown to likely drive shallow-marine biodiversity (Hannisdal and Peters, 2011). However, comparing the land-based and deep-sea microfossil records has shown that the relative availability of rock sections also shapes recorded diversity (Lloyd et al., 2012a, 2012b). Furthermore, no simple link exists between sea level, shallow-marine shelf area, and marine sedimentary rock area surviving as outcrop (Holland, 2012). Thus precisely how marine rock area, habitat diversity, and the fossil record of biodiversity are interrelated remains poorly understood.

Here we use the well-documented geological record of the United Kingdom (UK) and the recently revised taxonomic record of Cretaceous echinoids to quantify the relationship between sea level, marine shelf area, lithofacies diversity, surviving rock outcrop area, and sampled biodiversity during a single large-scale sea-level cycle.


The Cretaceous echinoid fauna of the UK was comprehensively revised (Smith and Wright, 1989, 1990, 1993, 1996, 1999, 2000, 2003, 2008, 2012), and comprises 228 species based on articulated tests; their stratigraphic distribution is drawn directly from the cited publications. The bedrock area of Cretaceous sedimentary rocks in England and Wales is estimated using the method in Smith and McGowan (2007). British Geological Survey maps at the scale of 1:50,000 provide a grid of equal-sized squares. The accompanying memoirs for each map were consulted to find the age of any Cretaceous bedrock that is, or has been, exposed as outcrop within each grid, and its lithology and environment of deposition. The British Geological Survey Lexicon of Named Rock Units (www.bgs.ac.uk/lexicon/) was consulted to obtain a standardized list of formation and member names applied to British Cretaceous rocks and their dominant sediment lithologies. A formation is a rock unit that is distinctive enough in appearance that a geologic mapper can tell it apart from the surrounding rock layers and thick enough and extensive enough to plot on a map at a given scale. Members are subdivisions of formations with distinctive lithologies that are defined independently of any fossil content, and represent discrete environmental settings. The eustatic sea-level curve, based on the sequence stratigraphy of Europe, comes from de Graciansky et al. (1998). Paleogeographic reconstructions for the UK in Hopson et al. (2010), Gale et al. (2000), and Ziegler (1990) were used to estimate the proportions of grid squares representing exposed massifs and flooded craton at selected times. While individual estimates come with some uncertainty, it is the emergent large-scale pattern of relative change over time that is important here. All data were compiled at the resolution of stage, giving 12 time bins, and are provided in the GSA Data Repository1. Pairwise statistical comparisons among time series were made using Spearman’s rank correlation of untransformed and first-differenced data. Multivariate models explaining echinoid species diversity were tested using generalized least squares regression incorporating an autoregressive model on the order of 0–3 in R version 2.10.1 (R Development Core Team, 2009), as explained by Benson and Butler (2011). Preferred autoregressive models and combinations of explanatory variables were selected using Akaike’s information criterion for finite sample size (AICc; Johnson and Omland, 2004).


Figure 1A shows the proportion of UK grid squares estimated to have been under fully marine conditions during each stage of the Cretaceous. It also shows the proportion of grid squares in which marine rock is now preserved as outcrop, and the relative proportion of original marine area captured by the rock record. The rock record is most complete during the early transgressive phase, but relative completeness drops sharply at maximum flooding and continues to decline during highstand.

Figure 1B shows the number of echinoid species recorded from the UK through the Cretaceous. Diversity is low in the early part of the Cretaceous and rises to a peak in the middle Cretaceous before falling sharply in the Turonian and again in the latest Cretaceous. Figure 1B also shows the bedrock area of marine sediments in the UK for the same time intervals. Bedrock area initially follows the diversity curve closely to reach a peak in the Turonian and then declines, gradually at first and then more steeply, through the Late Cretaceous.

Figure 1C shows the number of formally named marine lithofacies units in each time interval. This is low in the Early Cretaceous and rises to a peak in the Albian, after which it declines to a low in the Coniacian. Also shown are the relative proportions of four major lithofacies (sands, mudrocks, limestones, chalks and marls) through the Cretaceous. Figure 1C records the dominant sediment represented in each marine lithofacies unit present in a time bin. Only the Albian and Cenomanian capture a good representation of all four lithologies, and only chalks represent the Late Cretaceous. Figure 1C also plots the average bedrock area per formation for each stage (obtained by dividing the total grid square occupancy of bedrock for each interval by the number of lithofacies units). As water depth increases the average area covered by mappable lithofacies increases, as expected, with a peak in the Coniacian, when there are only two named formations.

Correlations among physical time series are given in Table 1, and show that bedrock area and lithofacies unit counts do not generally correlate with each other, especially after first differencing. Results of time series regressions are given in Table 2 and in the Data Repository. Bedrock area and lithofacies unit counts provide statistically significant, independent information on echinoid species diversity. Thus, a model including both explanatory variables has marginally the best AICc score, and both explanatory variables are statistically significant within this model. The Breusch-Pagan test indicates that other models with comparable AICc weights have heteroskedastic residuals, suggesting their support is overestimated. Sea level, lithofacies extent, and multiple regressions including these variables are poor explanations of species diversity (see the Data Repository). The excellent correspondence between a rock proxy, incorporating both bedrock area and lithofacies diversity, and sampled species diversity is apparent from Figure 2.


The geological record of the British Isles during the Cretaceous is that of passive margin subsidence created by the opening of the North Atlantic. As a result, the environments recorded in the rock record shifted from predominantly nonmarine and fluviatile-dominated to deep (200+ m) marine shelf, as the trailing edge sank faster than sediment accumulated (Mortimer et al., 2001; Hopson et al., 2010). There was a particularly marked flooding event toward the end of the Cenomanian, coinciding with the well-known Cenomanian-Turonian oceanic anoxic event (Gale et al., 2000). While this shift in water depth is clearly recorded in the changing nature of the sediments deposited and the marine environments they capture, the subsequent survival of these sediments in the geological record has been shaped by uplift and erosion associated with inversion structures generated by the northern push of Africa and later by the Alpine orogeny (Kley and Voigt, 2008). This began to affect the UK in the latest Cretaceous to early Paleogene and resulted in uplift of Cretaceous deposits, exposing them to erosion and stripping off the shallower and more recently deposited sediments. No marine Danian deposits (earliest Paleogene) occur in the British Isles.

The spatial extent of marine sediments and number of discrete lithofacies that survive at outcrop reflect the interplay of both deposition and erosion; the rise in outcrop area in the middle Cretaceous is a direct outcome of a major marine transgression, and the later Cretaceous progressive drop in outcrop area is the product of post-Cretaceous erosion (Fig. 1A). Lithofacies diversity peaks during transgressive systems tracts, while greatest geographical extent of formations and lowest lithofacies diversity occurs at maximum flooding and during early highstand, and shows no significant correlation to either sea level or bedrock area (Table 1). Furthermore, while the early transgressive phase results in a progressive rise in the proportion of original marine shelf area in the rock record, a major drop in the completeness of the sedimentary record occurs at the late Cenomanian flooding event, even though present-day outcrop area remains high (Fig. 1A).

The diversity of echinoids within this geological record is correlated with both the number of lithofacies units and the total area of rock available for sampling (Table 2). Because the great majority of echinoid species are restricted to a single lithofacies unit, better sampling within lithofacies units is likely to be less important in driving diversity patterns than better lithofacies unit sampling. While the flooding of the craton through the Cenomanian-Turonian led to a slight increase in total rock outcrop area, it also resulted in a marked drop in lithofacies diversity, and an equally marked drop in echinoid diversity. Outcrop area at such times is a poor predictor of lithofacies diversity, because flooding results in the deposition of near identical lithologies over wide areas. However, within a single lithology such as chalk, bedrock area affects the diversity of fossils, with net reduction in bedrock area resulting in a corresponding drop in recorded species numbers. A composite model comprising bedrock area and lithofacies units is the best predictor of sampled species diversity (Table 2), explaining ∼80% of the variance.

The recorded diversity of echinoids is part biological signal, as local communities respond to the environmental opportunities and challenges created by sea-level change, and part sampling bias imposed by an incomplete and heterogeneous geological record. However, the relative importance of these two drivers changes significantly through major sea-level cycles. During transgression, rising sea level creates increasing marine shelf area with more diverse lithofacies representation, and there is a parallel increase in the number of species known across the region. Echinoid diversity increased largely by range expansion, as few taxa are strictly endemic to the UK. The correlation between rock record and biodiversity during this phase thus fits the expectation of the common cause hypothesis (Peters, 2005). However, in the latter part of this sea-level cycle sampling biases become the dominant driver shaping the fossil record. At maximum flooding there is a sharp drop in both the proportion of original marine area that survives in the geological record and the variety of lithofacies preserved in that rock record, which becomes overwhelmingly dominated by deeper water deposits. While this clearly drives community structure locally, as deeper water environments become more widespread and specialist chalk faunas replace diverse, mixed shallow-water faunas, it also creates a marked bias in our ability to sample across the full range of shallow-marine habitats. This failure of the geological record to capture the shallower lithofacies that once presumably existed around the postulated land massifs must also contribute to the apparent decline in the recorded diversity of echinoids. A second drop in species diversity comes toward the end of highstand systems tracts, and coincides with a progressive loss of geological record. Here there is no change in lithofacies diversity, as the sedimentary record remains that of deep-water chalks unconformably overlain by post-Cretaceous deposits. However, exposures of younger sediments become fewer and progressively more restricted in outcrop area as a result of post-Cretaceous erosion, and there is a collateral degradation of the fossil record, with fewer collecting opportunities offered by the UK geological record leading to lower recorded diversities.

While our results pertain strictly to a single sea-level cycle in a single tectonic setting, a similar pattern in which outcrop area determines the number of fossil samples but not necessarily the diversity contained within those samples has been observed in the very different tectonic setting of New Zealand (Crampton et al., 2003). Diversity in the fossil record is thus not simply tied to the amount of bedrock available, but also to the diversity of habitats captured by that rock record. Furthermore, no simple invariant relationships exist between the amount of surviving rock area and lithofacies diversity and the original marine area and range of habitats; these are dependent on stratigraphic architecture.

Peters and Heim (2011) discovered an unexpected asymmetry to the relationship between macroevolution and macrostratigraphy. When they plotted the distribution of first and last occurrences of taxa against the distribution of originations and terminations of sediment packages they found that rates of last occurrence (their “extinction rates”) and sediment package truncation rates were significantly positively correlated, whereas rates of first occurrence (“origination rates”) were not correlated with sediment package origination rates. This relationship can now be explained: first and last appearances are indeed random in time. However, in the later stages of sea-level cycles erosional truncation of sediment packages and species ranges dominate, creating the strong correlation between last appearances and sediment package truncations. Critically, these truncations represent loss of information on the original geographic extent of sediments and range of lithofacies represented (i.e., sampling bias; Fig. 1A), rather than contraction of the original basin area (implied by the common cause hypothesis).

When estimating diversity, the ability to sample fairly across a broad range of habitats is as crucial as the ability to sample fairly within those habitats, but large-scale sea-level cycles clearly hinder the former. While the geological record left by a major sea-level cycle accurately captures the biological evolution of local communities, it is incomplete and biased in its coverage of environmental diversity. Only during major transgression intervals does the rock record capture a broad spectrum of lithofacies and their associated biotas. The ∼60 m.y. cycles reported in the sediment record (Meyers and Peters, 2011) and the 87Sr/86Sr strontium record (Melott et al., 2012) may well record a periodic pulse of the Earth, but it is by no means clear that the same cyclicity in sampled fossil taxa (Melott and Bambach, 2011) is primarily a biological signal.

We thank three anonymous referees and James Crampton for their insights and helpful criticisms of an earlier draft of this paper. Benson is supported by Leverhulme Foundation Research Project Grant 129.

1GSA Data Repository item 2013042, primary data used in this paper, is available online at www.geosociety.org/pubs/ft2013.htm, or on request from editing@geosociety.org or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301, USA.