Environmental determinants of marine benthic biodiversity dynamics through Triassic Jurassic time

Transcription

1 Paleobiology, 33(3), 2007, pp Environmental determinants of marine benthic biodiversity dynamics through Triassic Jurassic time Wolfgang Kiessling and Martin Aberhan Abstract. Ecology is thought to be of crucial importance in determining taxonomic turnover at geological time scales, yet general links between ecology and biodiversity dynamics are still poorly explored in deep time. Here we analyze the relationships between the environmental affinities of Triassic Jurassic marine benthic genera and their biodiversity dynamics, using a large, taxonomically vetted data set of Triassic Jurassic taxonomic occurrences. On the basis of binomial probabilities of proportional occurrence counts, we identify environmental affinities of genera for (1) carbonate versus siliciclastic substrates, (2) onshore versus offshore depositional environments, (3) reefs versus level-bottom communities, and (4) tropical versus non-tropical latitudinal zones. Genera with affinities for carbonates, onshore environments, and reefs have higher turnover rates than genera with affinities for siliciclastic, offshore, and level-bottom settings. Differences in faunal turnover are largely due to differences in origination rates. Whereas previous studies have highlighted the direct influence of physical and biological factors in exploring environmental controls on evolutionary rates, our analyses show that the patterns can largely be explained by the partitioning of higher taxa with different evolutionary tempos among environments. The relatively slowly evolving bivalves are concentrated in siliciclastic rocks and in level-bottom communities. Furthermore, separate analyses on bivalves did not produce significant differences in turnover rates between environmental settings. The relationship between biodiversity dynamics and environments in our data set is thus governed by the partitioning of higher taxa within environmental categories and not directly due to greater chances of origination in particular settings. As this partitioning probably has ecological reasons rather than being a simple sampling artifact, we propose an indirect environmental control on evolutionary rates. Affinities for latitudinal zones are not linked to systematically different turnover rates, possibly because of paleoclimatic fluctuations and latitudinal migrations of taxa. However, the strong extinction spike of tropical genera in the Rhaetian calls for an important paleoclimatic component in the end-triassic mass extinction. Wolfgang Kiessling and Martin Aberhan. Museum für Naturkunde, Humboldt-Universität, Invalidenstrasse 43, D Berlin, Germany. wolfgang.kiessling@museum.hu-berlin.de Accepted: 8 April 2007 Introduction There are several a priori reasons to expect biodiversity dynamics to be linked to the environment. First, common sense and theoretical considerations relate environmental settings with evolutionary rates. The amount of available energy in a particular environment and the temporal and spatial variability of the environment should affect the probabilities of speciation and extinction. Energy in the form of solar radiation or temperature could influence rates of genetic change (Davies et al. 2004; Evans and Gaston 2005; Allen et al. 2006); high-nutrient settings support larger populations, potentially reducing extinction risk (Parsons 1993, 2005); and both predation pressure and competition are thought to be greater in high-energy settings such as the Tropics (Vermeij 1987; Leighton 1999; Aberhan et al. 2006), potentially increasing escalation and thereby evolutionary rates. Environmental heterogeneity and disturbances at small temporal scales (stressed environments) tend to generate and maintain genetic and phenotypic variability and decrease gene flow (Hoffmann and Hercus 2000; Nevo 2001; Tainaka et al. 2006), potentially supporting speciation. Finally, environments that are frequently disturbed at geological time scales should exhibit higher rates of extinction than more stable environments, unless clades in these environments are adapted to these fluctuations (Sheldon 1996). Second, there is a rich body of empirical evidence that evolutionary novelty is linked to environments. For example, onshore settings (Jablonski et al. 1983; Bottjer and Jablonski 1988; Jablonski and Bottjer 1991; Jablonski 2007 The Paleontological Society. All rights reserved /07/ /$1.00

2 TRIASSIC JURASSIC DIVERSITY DYNAMICS ) and the Tropics (Jablonski 1993; Jablonski et al. 2006) appear to be loci of evolutionary novelty. It is a matter of debate whether the higher origination potential is evident only at higher taxonomic levels (Jablonski 2005), or whether it also plays out at the species level (Cardillo et al. 2005). The present study assesses whether the fossil record actually preserves systematic differences in the biodiversity dynamics of taxa with different environmental affinities. Our study builds on previous work that linked lithological affinities (also termed substrate affinities) to biodiversity dynamics (Miller and Connolly 2001; Foote 2006; Kiessling et al. 2007). A comprehensive set of environmental affinities has already been analyzed by Kiessling et al. (2007), who focused on extinction rates across the end-triassic mass extinction. We first identify the ecological affinities of genera for broadly defined environmental groups: lithology ( substrate), depositional environment ( bathymetry), community type (reef versus level-bottom), and latitudinal zone. We then assess the relationships between rates of extinction, origination, diversification, and turnover of genera and their ecological affinities, which are defined by the relative number of occurrences in a particular environment. We use a large, taxonomically vetted data set of Triassic through earliest Cretaceous benthic marine invertebrates. This time interval is especially interesting because it exhibits a great variability of diversity dynamics, incorporating the recovery period from the Permian-Triassic mass extinction, the end-triassic mass extinction, and less severe extinctions in the Early Jurassic and at the Jurassic/Cretaceous boundary. Methods Data. Species- and genus-level taxonomic occurrences from the Induan to the Barremian Stages were downloaded from the Paleobiology Database (PBDB, on 16 February The downloaded file included information on geology (lithology, depositional environment) and paleogeography (paleolatitude and paleolongitude), which are crucial to identify environmental affinities. Taxonomic data were revised asfar aspossible with the aim of identifying the correct genus assignments of all species occurrences in the data set, as described by Wagner et al. (2007). All non-benthic taxa (mostly cephalopods and fishes) and microfossils (foraminifera, conodonts, and ostracods) were excluded from the file prior to analyses. Genera with only one occurrence in the whole data set were also omitted, as were genera that could not be vetted and are very unlikely to occur in the Mesozoic. Corrections of stratigraphical assignments were largely limited to the Late Triassic following the redefinition of the Rhaetian Stage by Dagys and Dagys (1994). Accordingly, the Kössen Formation in Austria and the Gabbs Formation in the United States were reassigned to the Rhaetian instead of partially ranging into the Norian (see also Kiessling et al. 2007). In addition, although the Volgian is treated as equivalent to the Tithonian in the PBDB, the late Volgian is usually correlated with the Berriasian (Sey and Kalacheva 1999), so we made this correction in the downloaded data set. The final data set consists of 39,252 taxonomic occurrences of 1455 genera. The number of genus occurrences and genera varies considerably through time, suggesting a strong sampling bias (Table 1). Identifying Affinities. Our approach to identifying environmental affinities is similar to that of Miller and Connolly (2001), Foote (2006), and Kiessling et al. (2007). However, because our approach modified previous approaches in some ways, we present the complete schema herein. The aim of the environmental categorization was a bipartite subdivision of environmental affinities of taxa within four environmental categories. We contrasted occurrences in (1) carbonate versus siliciclastic lithologies ( substrates), (2) onshore versus offshore depositional environments ( bathymetry), (3) reef versus levelbottom community types, and (4) tropical versus non-tropical latitudes. These variables are obviously not independent. For example, occurrences in carbonate lithologies may largely overlap with occurrences in tropical latitudes. However, it is sensible to treat the environmental variables in isolation first and then discuss possible contingencies. Occurrences without a clear identification of lithol-

3 416 WOLFGANG KIESSLING AND MARTIN ABERHAN TABLE 1. Summary of raw data by stratigraphic interval (bin). Interval Abbrev. Age at base (Ma) Taxonomic occurrences Sampled genera* Boundary-crossing genera Early Triassic SCYT Anisian ANIS Ladinian LADI Carnian CARN Norian NORI Rhaetian RHAE Hettangian HETT Sinemurian SINE Pliensbachian PLIE Toarcian TOAR Aalenian AALE Bajocian BAJO Bathonian BATH Callovian CALL Oxfordian OXFO Kimmeridgian KIMM Tithonian TITH Berriasian BERR Valanginian VALA Hauterivian HAUT * Including genera limited to one bin. ogy or depositional environment were discarded for the determination of affinities of a genus but were used for the assessment of generic biodiversity dynamics if their affinities could be identified by other occurrences of the same genus. The lithological assignments relied on the primary lithology of a collection of fossils as given in the database. We ignored the secondary lithology, as well as all collections classified as marls or mixed carbonate-siliciclastic. A full list of lithologies is given by Foote (2006). We used the depositional environment field in the database to categorize both bathymetry and community type. Onshore-offshore sediments were separated by whether they were deposited above or below the storm wave base and were classified as detailed by Kiessling et al. (2007). Although the great majority of Triassic Jurassic reefs grew above storm wave base (Kiessling et al. 1999), we excluded reefs from the onshore-offshore categorization because the bathymetric setting of reefs is not recorded in the PBDB and we wanted to avoid an a priori mixture of bathymetry and community type. Reefal occurrences were identified if the collection record had the term reef, buildup or bioherm in the depositional environment field. All other depositional environments were classified as level-bottom communities, excluding collections where the environment was unidentified or only broadly characterized ( marine indet. and carbonate indet. ). The assignment of occurrences to paleolatitudes relied on the plate tectonic rotation files, which are based on Scotese (2001) and implemented into PBDB s download scripts. The demarcation of tropical versus non-tropical occurrences was arbitrarily defined at 30 paleolatitude, which results in approximately equal occurrence counts in either category. This static demarcation of latitudinal zones does not reflect climatic fluctuations but may serve as a rough proxy of climatic setting. The procedure to identify environmental affinities of genera consists of four steps: First, the occurrence counts of each genus were summed for each environmental setting. Second, the stratigraphic range of each genus was determined from the raw data in PBDB. Third, all occurrences in each environmental category were summed (N t ) over the stratigraphic range of this genus, to evaluate the sampling intensity of environments. Fourth, the environmental affinity of each genus was identified by using binomial probabilities of apparent affinities. The apparent affinity of genus g for one of two environmental settings (h, i)

4 TRIASSIC JURASSIC DIVERSITY DYNAMICS 417 was determined by proportional occurrence counts: Nght Ngit Ag (1) N N where t denotes the stratigraphic range of genus g. If A g is greater than 1, then the genus would be classified as having an apparent affinity for setting h, and if it is less than 1 the apparent affinity would be for setting i. Whether A g is significant was determined by the binomial probability of recording k or more occurrences of genus g in one setting given N ht and N it. The binomial probability P for drawing exactly k out of n occurrences given p and q ( 1 p) is n! k n k P (P )(q ). (2) k!(n k)! If settings h and i were equally sampled, then p would be 0.5. However, one type of environmental setting is usually more commonly represented by fossils than the other, and thus the value of p is larger or smaller than 0.5. We illustrate this method with an example from our data set. The Jurassic gastropod Bourguetia has 43 occurrences, 6 of which are in reefs (h), 25 in level-bottom communities (i), and 12 in unspecified community types. From these counts alone, it appears that Bourguetia is more likely to occur in level-bottom communities. However, reef communities are much rarer than level-bottom communities. In the raw life span of Bourguetia (Hettangian to Kimmeridgian), there are 2153 reefal occurrences and 10,975 level-bottom occurrences in the data set. Thus the true probability p of recording the genus in reefs, if it did not have any affinity, would be 2153/( ,975) A g in this example is (6/2153)/(25/ 10,975) 1.22, which suggests an apparent affinity for reefs. But this preference is far from being significant. The binomial probability of finding six or more occurrences of Bourguetia in reefs given the general proportions of occurrences is 0.427, suggesting that the snail was a generalist with respect to community type. We used the number of total occurrences in ht tt each environment to assess sampling probability because we did not combine multiple occurrences within a collection. However, the method yields the same results if the number of collections is used to assess sampling. With this new approach it is not necessary to use an arbitrary lower bound to identify environmental affinities (as used by Foote [2006] and Kiessling et al. [2007]). Ideally, we would want to consider only genera with a significant affinity for one type of setting, that is, those that have a binomial probability of 0.05 that the affinity is due to chance. However, in some analyses, this level of significance excludes too many genera and occurrences to achieve adequate analyses (see results). In recognizing that the absence of a strictly significant affinity is largely due to low sample sizes, we have therefore also used a higher threshold for assigning affinities (p 0.1), which we refer to as marginally significant. The example of Bourguetia has already shown that in addition to identifying taxa with environmental affinities, we are also able to identify taxa without a significant affinity, that is, taxa that are generalists with respect to particular environmental categories. However, not all genera that fail to exhibit significant affinities can be referred to as generalists. There are two possible reasons why taxa would not show a significant affinity. One is purely statistical, that is, the number of occurrence counts is insufficient to reach a significant threshold, even if all occurrences of a genus were from one setting only. The other has biological meaning: there would be sufficient occurrences for reaching a significant threshold, but this threshold is not reached. Only if both of the possible minimum binomial probabilities (all occurrences in setting h and all occurrences in setting i) could both reach a significant threshold is a biological explanation likely. For the purpose of this paper, we have qualified as generalists only those taxa that could have reached a significant threshold given their abundance, but whose actual probability of having an affinity to a particular setting is not significant. Diversity Dynamics. Our number of genera is less than half of the 3280 macrobenthic genera recorded in Sepkoski s compendium (Sep-

5 418 WOLFGANG KIESSLING AND MARTIN ABERHAN koski 2002) for the investigated time interval. However, the pattern of extinction rates calculated from PBDB raw data matches closely the rates calculated from Sepkoski s compendium except for the more pronounced Rhaetian spike in the PBDB, which is probably due to the late addition of the Rhaetian Stage in Sepkoski s data set (Kiessling et al. 2007). Diversity dynamics were assessed at the genus level and for standard stratigraphic stages (except for the Early Triassic, where the Induan and the Olenekian Stages were always combined to achieve a sufficient sample size). The origination and extinction rates reported here are strictly based on sample-standardized analyses. We randomly drew the same number of taxonomic occurrences (subsampling quota) from the pooled data set without replacement and ranged the data through to evaluate diversity dynamics ( simple rarefaction of Alroy et al. 2001). This procedure of random draws and range-through analysis was repeated 100 times for each analysis (subsampling trials or iterations) and the results were averaged. Extinction (E) and origination (O) rates are reported as per-genus rates following Foote (2000, 2003). Turnover rates (E O) and diversification rates (O E) were also computed. In accordance with Foote (2003, 2005) we did not normalize our measures for the duration of intervals. Given the limited stratigraphic range of our analyses, there are profound edge effects with this counting method; that is, origination rates become systematically larger toward the beginning of the time series and extinction rates are artificially inflated toward the end of the time series. However, these edge effects are unlikely to affect the comparison of diversity dynamics, which is the aim of our analysis. In the graphs below, we dampen the edge effect by depicting origination rates only from the third interval onwards and extinction rates only until the third interval backwards. Because in most cases the occurrence counts are approximately equal between environments (see Results ), we used the same subsampling quota for most pairwise comparisons within categories. However, the much smaller number of occurrences of reef genera calls for a higher quota in level-bottom occurrences (see Appendix). Lumping of stages was necessary in several analyses to achieve an adequate subsampling quota (see results). The lumping represents a tradeoff between the preservation of as many time intervals as possible and a sound subsampling quota. Even with lumping, the quota often had to be quite low to preserve a complete time series. This rigorous treatment of the data may appear too harsh. However, sampling standardization is crucial to achieving meaningful patterns, because the number of occurrences of taxa with particular environmental ( ecological) affinities is very volatile, as is the preservation of environments itself (Peters 2005). With sampling standardization, we can focus on data present in the database and be less concerned about data that may still be missing. Adding more collections to bins that are already well sampled (e.g., the Late Triassic and Late Jurassic for all environmental settings) does not help to improve the accuracy of our analysis because the most poorly sampled bins determine the subsampling quota. Some of these poorly sampled bins are not just due to an incomplete coverage of the published literature in the PBDB, but rather reflect the nature of the geological record and perhaps ecological changes. For example the well-known metazoan reef gap in the Early Triassic and the very rare reefs recorded in the Hettangian are probably due to a collapse of reef building in the aftermath of mass extinctions (Flügel and Kiessling 2002). Results Distribution of Ecological Affinities. Ecological affinities vary greatly between taxonomic groups, environmental categories, and stratigraphic intervals. The majority of genera are too rare to determine an ecological affinity. Of our 1455 genera occurring in more than one collection, only 1076 could exhibit a significant affinity for any of the four environmental types if all their recorded occurrences were in one setting. Of these assignable genera, 869 have a significant affinity for at least one environmental category. Most of these genera are common, so that this 40% reduction in ge-

6 TRIASSIC JURASSIC DIVERSITY DYNAMICS 419 TABLE 2. Counts of genera and their environmental affinities among common higher taxa (classes and phyla) with at least 1000 taxonomic occurrences in the data set and among all taxa. Class/Phylum Genera Carbonates; siliciclastics No. of genera with a significant (p 0.05) affinity for Onshore; offshore Reefs; level-bottom Tropics; non-tropics Bivalvia ; ; 57 16; ; 100 Brachiopoda ; 15 27; 24 6; 39 38; 52 Anthozoa ; 9 31; 0 106; 8 21; 29 Gastropoda ; 64 16; 30 8; 28 30; 55 Echinodermata ; 11 1; 11 4; 5 20; 16 Porifera ; 1 1; 1 59; 0 25; 1 All groups ; ; ; ; 270 nus counts translates into a reduction of occurrence counts by only 7.3%. Environmental affinities are nonrandomly distributed among higher taxa. Table 2 lists genus counts with significant affinities for higher taxa with at least 1000 occurrences in the data set. Bivalves and gastropods predominantly have siliciclastic (named clastic for simplicity) affinities, whereas other taxa tend to have affinities for carbonates. The affinity of higher taxa is about equally distributed between onshore and offshore settings, except for corals, which have a strong affinity for onshore settings. Similarly, corals and sponges have strong affinities for reefs, whereas all other common taxa are mostly linked to levelbottom ecosystems. Only sponges exhibit a strong preference for tropical latitudes, whereas most genera in other higher taxa appear to be more strongly linked to non-tropical latitudes. The total number of genera with significant affinities is approximately equal for all but one environmental category: only 268 genera can be significantly assigned to either an onshore or offshore affinity, whereas 447 to 562 genera have significant affinities in the other categories. Are particular environmental categories more likely to host genera with significant affinities than others, and is the proportion of these taxa different within categories? To address these questions, one first needs to consider the different sample sizes on which the affinity assignments were made. Because all data in the database have geographic coordinates, they all are rotated with Scotese s (2001) rotation file (100%). Therefore, the assignment of latitudinal affinities is based on the largest amount of data. The primary lithology is also noted for the great majority of occurrences (95%), but as we excluded mixed lithologies, this number drops to 83%. Only 58% of the occurrences could be reliably assigned to reef versus level-bottom communities and for just 43% of the occurrences could onshore-offshore settings be separated. These differences in sample size affect the minimum possible binomial probability and thus the number of possible affinity assignments. In our four environmental categories, the proportions of genera that demonstrated significant affinities (relative to all those that could have) are usually large and do not vary much (Fig. 1A). The affinities to latitudinal zones, however, represent an outlier, with just 59% of the genera being significant. This suggests that the overall environmental affinity of genera is weakest for latitudinal zones, which could be explained by migrations of taxa with climatic change. The affinities of genera for settings within the environmental categories are fairly uniform but there are two significant differences (Fig. 1B): The proportion of genera with an affinity for carbonates is greater than for clastic substrates (p 0.03, two-tailed t-test) and for level-bottom communities than for reefs (p 0.04). This might indicate that it is easier for taxa to move from clastic substrates to carbonates and from reefs to level-bottom communities than the other way. The raw occurrence counts within specific environments vary strongly through time, even when proportional values are considered (Fig. 2A,B). This volatility (measured by the standard deviation of first differences) is strongly reduced for proportions of genus oc-

7 420 WOLFGANG KIESSLING AND MARTIN ABERHAN FIGURE 1. Proportions of genera with significant affinities for environmental categories (A) and settings within these categories (B). A, Proportions of genera that do reach a significant threshold within environmental categories relative to all taxa that could reach this threshold given their abundance and the sampling of environments. The likelihood for significant affinities is substantially lower for latitudinal zones than for the other categories. B, Proportions of genera with significant affinities for particular settings within categories. Note the greater proportions for carbonate than for siliciclastic substrates and for level-bottom communities than for reefs. Error bars in this and the subsequent figures are plus and minus one standard error. currences with significant affinities for the same environments (Fig. 2C,D), although the variations of affinity counts are still substantial. Short-term fluctuations of environmental information or of the paleontological record of environments are balanced by good information in other intervals. An example is the Hettangian, which has very few reef occurrences owing to the near absence of reefs in this stage (Flügel and Kiessling 2002). However, the majority of Hettangian corals have reef affinities over their entire stratigraphic range, even though they are mostly found in level-bottom ecosystems in this stage. Trajectories of Extinction and Origination Rates. The overall trajectories of extinction and origination rates (Fig. 3) suggest two major extinction spikes, one at the end-triassic (Rhaetian) and one at the end-jurassic (Tithonian). The end-jurassic extinction has to be viewed with caution, because of edge effects and the change in sampling regimes (Kiessling and Aberhan 2007). Background extinction levels are comparatively high in the Triassic. Bearing in mind that we did not normalize for interval duration, this might be due to the fairly long duration of many Triassic stages. However, both extinctions and originations tend to be pulsed within stages rather than gradually accumulating within stages (Foote 2005), and we do not find a significant correlation between turnover metrics and interval duration in our data set, even if we exclude the two prominent extinction spikes. Origination rates show pronounced spikes in the Hettangian, Bajocian, and Valanginian. The large values in the pre-rhaetian Triassic are probably all related to the recovery from the end-permian mass extinction. The other high values in origination rates can also be attributed to the recovery of extinctions (Hettangian and probably Valanginian), but the Bajocian spike is not obviously linked to a previous extinction. In the following sections, we compare trajectories of extinction and origination rates for taxa with significant affinities for different settings. When contrasting results become apparent with a lower threshold, we also report patterns based on marginally significant affinities. Substrate. Both extinction and origination rates are higher on average for genera with a significant affinity for carbonates than for genera with a preference for clastics (Fig. 4). A Wilcoxon signed ranks test, however, indicates that only origination rates are significantly larger over the entire time series (p 0.002) (Table 3). The lack of significance for the differences of extinction rates is due to a marked change in the Middle Jurassic. In the Middle Triassic to Early Jurassic, extinction rates were always larger for carbonate dwellers and even significantly so in the Ladinian and Carnian Stages (Fig. 4). Extinction rates converged in the Middle and Late Jurassic. In the Callovian, there even was a significantly larger extinction rate in clastic than in carbon-

8 TRIASSIC JURASSIC DIVERSITY DYNAMICS 421 FIGURE 2. Time series of proportional occurrence counts by environmental category. The values refer to the percentages of occurrences in one of the paired settings relative to all occurrences in the category. For example, the percentages of carbonate occurrences represent: carbonate occurrences/(carbonate occurrences clastic occurrences) 100. A and B, Raw occurrence percentages. C and D, Occurrence percentages of genera with significant affinities. The drop of several counts in the Aalenian (AALE) is due to prevalence of collections from non-tropical, offshore, clastic settings in this stage. Consequently, the Aalenian has been combined with the Toarcian in nearly all analyses. See Table 1 for the abbreviation of stage names. ate dwellers. Origination rates, however, remained larger for carbonate dwellers in the Middle and Late Jurassic, being significantly larger in the Carnian, Toarcian Aalenian, Bajocian, Oxfordian, and Kimmeridgian. As a result, the diversification rates through the entire time series are significantly larger for carbonate dwellers (Table 3). Diversity dynamics are strongly cross-correlated between the two settings (Table 4). This suggests a common trigger of evolutionary change, whereby genera associated with carbonates respond generally more sensitively to this trigger than genera associatedwithclastics. Theresultsare stable, with varying probability thresholds for affinities between 0.01 and 0.1. Bathymetry. Genera associated with onshore settings do not generally exhibit greater extinction rates than genera with offshore affinity, but there is a tendency for origination rates to be larger onshore than offshore (Fig. 5A). Turnover and diversification rates do not differ significantly between onshore and offshore affinities. The overall difference in origination rates, although significant (Table 3) is arguably weak. Because sample size is an issue for bathymetry (see section on distribution of ecological affinities and Table 2), a lower threshold for defining affinities is especially instructive. When marginally significant affinities are included in the analysis, the number of genera increases by 24% to 331. This addition has a marked effect on the outcome (Table 3, Fig. 5B). The difference in origination rates becomes larger and turnover rates are now also significantly higher for onshore genera. Although the difference in extinction rates is now markedly larger in the Late Triassic, there is still no general tendency for higher extinction rates onshore. Similar to the results on substrate affinities, the differences in rates are higher in the Triassic than in the Jurassic. The cross-correlation of rate trajectories is somewhat lower than for substrate affinities but still significant for all metrics (Table 4). Community Type. The trajectories of extinction and origination rates for community types differ from the previous patterns in that

9 422 WOLFGANG KIESSLING AND MARTIN ABERHAN FIGURE 3. Time series of sample-standardized extinction and origination rates of benthic marine organisms from the Triassic to the earliest Cretaceous. Trajectories in this and subsequent figures are the average of random draws of a subset of occurrences in each time interval (bin) without replacement and an analysis of stratigraphic ranges after 100 iterations. They report the metrics of Foote (2000, 2003). Error bars in this and subsequent figures give plus and minus one standard error, which in equivalence to bootstrap is the standard deviation of mean rates achieved in all subsampling trials. The Triassic/Jurassic boundary and the Jurassic/ Cretaceous boundary are marked by dashed vertical lines. Subsampling quota: 720 occurrences in each bin. FIGURE 4. Trajectories of extinction and origination rates of all benthic genera with a significant (p 0.05) affinity for carbonate and clastic ( siliciclastic) substrates. Extinction rates of carbonate dwelling taxa tend to be higher in the Triassic and Early Jurassic but the difference is not significant over the entire time series. Origination rates, however, are significantly elevated in carbonates, although there are few exceptions. Note that the Toarcian and Aalenian Stages were combined to reach an adequate subsampling quota of 115 occurrences per bin and substrate category. extinction rates diverge more strongly between settings (Fig. 6), although the overall extinction rates of reef taxa are not significantly higher than for level-bottom taxa at an affinity threshold of The higher reefal extinction rates are especially pronounced in the Rhaetian and Pliensbachian. Origination rates are significantly higher for reef genera than for level-bottom genera over the entire time series (Table 3). This difference is largely due to the much higher rates in the Norian and most of the Jurassic. A prominent exception in the Jurassic is the smaller origination rate of reef genera during the Bajocian. Origination rates of reef genera declined in this stage, whereas most genera show a pronounced origination peak (Fig. 3). Turnover is also significantly higher in reefs but diversification is not. Another interesting difference from the previous results is the lack of correlation between changes in origination rates. Whereas extinction, turnover, and diversification rates are significantly cross-correlated, origination rates of reef and level-bottom communities appear to be completely decoupled (Table 4). This may suggest that extinctions of reef taxa are driven by the same, probably physicochemical factors as extinctions in level-bottom taxa, but origination rates are triggered by different factors. Because the trajectories of origination rates of level-bottom genera are more similar to the trajectories of the other settings, these different factors are likely to prevail in reefs.

10 TRIASSIC JURASSIC DIVERSITY DYNAMICS 423 TABLE 3. Probabilities that the observed differences between the diversity dynamics of taxa with contrasting affinities for particular environments are due to chance (based on Wilcoxon signed ranks test). Significant values are in bold. Environmental pair Probability threshold for affinities Extinction rate Origination rate Turnover rate Diversification rate Carbonate clastic Carbonate clastic Onshore offshore Onshore offshore Reef level-bottom Reef level-bottom Tropical non-tropical Tropical non-tropical Even with lumping of the Early Triassic Anisian, the Toarcian Aalenian and the first three stages of the Cretaceous, the maximum subsampling quota of the analysis in Figure 6 was 75 occurrences per bin and setting, because the Hettangian has just 76 significant occurrences of genera with reef affinities. This quota is arguably very low. We thus ran another analysis in which the Hettangian was combined with the Sinemurian, permitting a quota of 120 occurrences for reef genera. The results are basically identical, with origination rates and turnover rates significantly higher for reef taxa than for level-bottom genera (p and p 0.006, respectively) but no significant differences in extinction and diversification rates. The total number of genera in reefs and level-bottom communities is similar, but occurrences of level-bottom genera are four times more common than occurrences of reef genera. Following our modeling results (see Appendix), we have thus chosen a higher subsampling quota for level-bottom genera. However, the basic results are identical if the same subsampling quota is used for both community types. Latitude. Affinities for latitudinal zones are least related to turnover rates. Neither extinction nor origination rates are consistently higher in either setting (Table 3, Fig. 7). Even with marginally significant affinities, there is no tendency for tropical genera to show larger extinction or origination rates than non-tropical genera. This result could have been expected, given that our separation of tropical and non-tropical zones is artificial and static. We would be better off if temporal expansions and contractions of the tropical zone could be considered. However, this is currently not feasible because the global climatic variability during Triassic Jurassic times is still poorly resolved, although a wealth of data has been gathered (e.g., Simms and Ruffell 1990; Vakhrameev 1991; Price and Sellwood 1994; Balog et al. 1999; Dromart et al. 2003; Gröcke et al. 2003; Ziegler et al. 2003; Rees et al. 2004; Rosales et al. 2004; Sellwood and Valdes 2006). Only one feature in the trajectories is prominent: the much more pronounced extinction of tropical than non-tropical genera in the Rhaetian, implying a paleoclimatic cause of the end-triassic mass extinction. The origination rate in the Anisian (not shown) and Ladinian is much larger for tropical genera TABLE 4. Cross-correlations of biodiversity dynamics of genera with significant affinities within environmental categories (based on Spearman-rank correlations of first differences). All correlations are significant at p 0.05, except the ones marked with (n.s. not significant). Environmental pair Extinction rates Origination rates Turnover rates Diversification rates Carbonate clastic Onshore offshore* Reef level-bottom (n.s.) Tropical non-tropical (n.s.) * Results based on marginally significant affinities.

11 424 WOLFGANG KIESSLING AND MARTIN ABERHAN FIGURE 5. Trajectories of extinction and origination rates of all benthic genera with affinities for onshore and offshore settings. A, Trajectories for significant affinities (p 0.05) with a subsampling quota of 100 occurrences. B, Trajectories for marginally significant affinities (p 0.1) with a subsampling quota of 120 occurrences. Origination rates tend to be higher onshore than offshore. than for non-tropical genera, but in the Carnian the origination rate is significantly lower. Thus recovery from the end-permian mass extinction may have been a largely tropical phenomenon, but there is no general tendency of higher origination rates in the Tropics. The first differences of extinction, turnover, and diversification rates are significantly crosscorrelated for tropical and non-tropical genera (Table 4). Although the correlation coefficients are somewhat smaller than for substrates and bathymetry, these correlations suggest that global physico-chemical changes were important in controlling biodiversity dynamics, because the genera are geographically separated. Environmental Generalists. Genera that are sufficiently common to theoretically reach a significant threshold but fail to do so, are regarded here as generalists with respect to environmental category. The comparison of diversity dynamics between specialists (genera with significant affinities) and generalists serves as a test for the role of affinity per se in determining turnover rates. Because further combination of stages was necessary for reaching meaningful quotas, the resulting time series are based on very few data, which renders it difficult to achieve significant results. We focus here on the three environmental categories that produced significant differences in one or more rate metrics, that is, substrate, bathymetry, and community type (Fig. 8). For substrates and community types, we contrasted rates of genera having significant affinities (p 0.05) with rates of genera that have no strictly significant affinities (p 0.05) but could have reached significant values given their abundance. For bathymetry, we applied the same rationale but used marginally significant affinities as a threshold (p smaller or larger than 0.1). The extinction rates of substrate generalists are not systematically different from the rates

12 TRIASSIC JURASSIC DIVERSITY DYNAMICS 425 FIGURE 6. Trajectories of extinction and origination rates of all benthic genera with a significant (p 0.05) affinity for reefs and level-bottom communities. As genera with reef affinity are much rarer than genera with level-bottom affinity whereas diversities are similar, a higher subsampling quota was chosen for level-bottom affinities (200 occurrences) than for reef affinities (75 occurrences). See Appendix for a justification of this approach. There is a significantly higher origination rate of reef genera than of level-bottom genera. FIGURE 7. Trajectories of extinction and origination rates of all benthic genera with a significant (p 0.05) affinity for tropical ( 30 paleolatitude) and non-tropical latitudes ( 30 paleolatitude). There is no significant difference in diversity dynamics. Note, however, the strong Rhaetian extinction spike of tropical genera. Subsampling quota: 130 occurrences. of specialists (Fig. 8A). Origination rates of generalists are usually situated between carbonate and clastic rates but both origination and turnover rates of carbonate specialists are significantly higher over the entire time series (p and p 0.009, respectively). The patterns for generalists with respect to bathymetry are quite different (Fig. 8B). Here extinction rates are significantly lower for generalists than for both onshore and offshore specialists (p and p 0.046, respectively), but there is no systematic difference in origination and turnover rates. Generalists with respect to community type do not exhibit systematically different turnover rates than specialists (Fig. 8C). Although Jurassic origination rates of generalists tend to be lower than origination rates of reef specialists, their rates are higher in two Triassic bins (Carnian and Norian) so that the overall difference is not significant (p 0.064). In summary, generalists show a tendency for reduced turnover only when compared with the more sensitive type of specialists. Turnover rates of generalists are thus not universally lower than turnover rates of specialists. Because generalists probably have broader ecological niches than specialists, one is tempted to reject the hypothesis that niche breadth per se is a major determinant of longevity (Kammer et al. 1997, 1998; Fernandez and Vrba 2005). However, abundance patterns may override the role of habitat specificity (Reinhardt et al. 2005), and abundance, measured by the number of occurrences, is an im-

13 426 WOLFGANG KIESSLING AND MARTIN ABERHAN FIGURE 8. Comparison of extinction and origination rates of genera with (specialists) and without (generalists) affinities for environments. Generalists are only those taxa that could have reached a significant (or marginally significant) affinity given their abundance but fail to do so. A, Trajectories for genera with affinities and for generalists with respect to substrates, applying a p 0.05 cutoff. Subsampling quota 110 occurrences. B, Trajectories of genera with affinities and for generalists with respect to bathymetry applying a p 0.1 cutoff. Subsampling quota 125 occurrences. C, Trajectories of genera with affinities and for generalists with respect to community types applying a p 0.05 cutoff. Subsampling quota 75 occurrences for reef affinities and 100 occurrences for both level-bottom affinities and generalists. Generalists do not generally exhibit lower turnover than specialists. Only in the case of bathymetric settings are generalists less prone to extinction than specialists in either onshore or offshore settings. portant determinant of genus longevity in our data set (Kiessling and Aberhan 2007). Generalists are usually rarer than specialists (on average 36 occurrences per genus for specialists and 27 occurrences for generalists), which may partly explain their relatively high turnover rates. Approaching Underlying Factors Comparison of Stratigraphic Ranges. We have deliberately focused on diversity dynamics rather than stratigraphic ranges until now. The dynamic assessment should be favored over simple estimates of stratigraphic ranges, because temporal variations of background extinction rates may bias the results. Mass extinctions truncate ranges, whereas unusually low background extinction rates tend to extend ranges. Are the different turnover rates within environmental categories reflected by differences of stratigraphic ranges? To answer this question, we have compared the mean values of the sampling standardized stratigraphic ranges (quota of 720 occurrences, single occurrences omitted) between settings. From our previous results on turnover rates (Table 3), we would expect genera with carbonate, onshore, or reef affinities to have shorter ranges than genera with clastic, offshore, or level-bottom affinities. However, Mann-Whitney U-tests confirm only that reef genera exhibit significantly shorter stratigraphic ranges than level-bottom genera (p 0.001). The absence of significant differences for substrates and bathymetry can be explained by the stronger dependency of turnover rates on origination rates; that is, changes of turnover rates are better correlated with changes of origination rates than with changes of extinction rates. The higher turnover of genera with carbonate and onshore affinities therefore is not reflected in shorter stratigraphic ranges. Separation of Affinities. Several of our environmental categories are obviously not independent. For example, carbonates are more widespread in the Tropics and in shallow water than outside the Tropics and offshore, at least prior to the widespread sedimentation of pelagic carbonate by calcareous plankton

14 TRIASSIC JURASSIC DIVERSITY DYNAMICS 427 TABLE 5. Contingency coefficients between significant affinities. All coefficients are significant at p Environmental category Community Substrate Bathymetry type Substrate Bathymetry 0.37 Community type Latitude (Walker et al. 2002). Cross-tabulations confirm that there are significant contingencies among all affinities in environmental categories (Table 5). Contingencies were determined for pairs of categories and genera that have significant affinities within both categories. Expectedly, the contingency coefficient is largest between community type and substrate, as reefs are carbonate substrates and all genera except one (Indopecten) that have significant affinities for reefs also have significant affinities for carbonates. The contingencies are moderate between bathymetry and latitude (onshore affinities coincide with tropical affinities, offshore affinities with non-tropical affinities); substrate and latitude (carbonate-tropical, clastic-non-tropical affinities); substrate and bathymetry (carbonate-onshore, clastic-offshore affinities); and bathymetry and community type (level-bottom-offshore affinities). The weakest contingency is between community type and latitude. Although the majority of genera with reef affinity also have a tropical affinity and level-bottom affinities tend to be concentrated in higher latitudes, the contingency coefficient indicates just 25% overlap. To assess underlying causes of differential turnover rates, affinities could be analyzed separately (e.g., carbonate but not reef) or in combination (e.g., carbonate and onshore versus clastics and offshore). We focus here on those patterns which yielded significant results in the individual analyses: substrate, bathymetry and community type. Because sample sizes are strongly affected by both separation and combination, we only report results from analyses using a probability threshold of 0.1 for assigning affinities. In addition to differences in nutrient regimes (see Discussion ), the carbonate-clastic difference in turnover rates could reflect a paleoclimatic signal, because marine carbonates are more likely to be deposited under warm, dry climates than under cool, wet conditions (Frakes et al. 1992; Riding 1993; Kiessling et al. 2003). The climate hypothesis is difficult to test because the separation of latitudinal zones did not produce significant results (see section on latitude). The four combinations of substrate and latitudinal affinities did not exhibit strong differences in turnover rates, except that genera with nontropical and carbonate affinities have significantly higher rates than either genera with non-tropical and clastic affinities (p 0.016) or genera with tropical and clastic affinities (p 0.033). These results cannot be taken as evidence that climate was not a major trigger of the carbonate-clastic difference in biodiversity dynamics, but they do suggest that additional factors may be important. Could the inclusion of reef taxa in the carbonate category drive the pattern? To test this, we have excluded all taxa that have a significant reef affinity from the analysis (Fig. 9). Extinction, origination, and turnover rates are significantly higher for nonreefal carbonate dwellers than for non-reefal clastic dwellers (p 0.02, 0.006, and 0.001, respectively). This suggests that the carbonateclastic difference in diversity dynamics is not driven by reefs. Does the pattern also hold for different bathymetric settings? Owing to a strong reduction of sample size, the combination of substrate and bathymetric affinities works only with lumping of several stages and low subsampling quotas. Even then it can be analyzed only for three combinations: carbonate and onshore, clastic and offshore, and clastic and onshore. The absence of any significant differences in biodiversity dynamics among these combinations does suggest that different factors, perhaps acting in different directions, are behind the differential diversity dynamics of substrate and bathymetric affinities. We have excluded reefs from the assignment of bathymetric affinities, which a priori separates bathymetry and community types. However, when treating all reefs as onshore, the differences of biodiversity dynamics remain strong, with significantly higher origination and turnover rates onshore than offshore (p 0.016