Dissecting the Paleocontinental and Paleoenvironmental Dynamics of the Great Ordovician Biodiversiﬁcation

Paleobiology, 45(2), 2019, pp. 221–234 DOI: 10.1017/pab.2019.4

Dissecting the paleocontinental and paleoenvironmental dynamics of the great Ordovician biodiversiﬁcation

Franziska Franeck and Lee Hsiang Liow

Abstract.—The Ordovician was a time of drastic biological and geological change. Previous work has suggested that there was a dramatic increase in global diversity during this time, but also has indicated that regional dynamics and dynamics in specific environments might have been different. Here, we contrast two paleocontinents that have different geological histories through the Ordovician, namely Laurentia and Baltica. The first was situated close to the equator throughout the whole Ordovician, while the latter has traversed tens of latitudes during the same time. We predict that Baltica, which was under long-term environmental change, would show greater average and interval-to-interval origination and extinction rates than Laurentia. In addition, we are interested in the role of the environment in which taxa originated, specifically, the patterns of onshore–offshore dynamics of diversification, where onshore and offshore areas represent high-energy and low-energy environments, respectively. Here, we predict that high-energy environments might be more conducive for originations. Our new analyses show that the global Ordovician spike in genus richness from the Dapingian to the Darriwilian Stage resulted from a very high origination rate at the Dapingian/Darriwilian boundary, while the extinction rate remained low. We found substantial interval-to-interval variation in the origination and extinction rates in Baltica and Laurentia, but the probabilities of origination and extinction are somewhat higher in Baltica than Laurentia. Onshore and offshore areas have largely indistinguishable origination and extinction rates, in contradiction to our predictions. The global spike in origination rates at the Dapingian/Darriwilian boundary is apparent in Baltica, Laurentia, and onshore and offshore areas, and abundant variability in diversification rates is apparent over other time intervals for these paleocontinents and paleoenvironments. This observation hints at global mechanisms for the spike in origination rates at the Dapingian/Darriwilian boundary but a domination of more regional and local mechanisms over other time intervals in the Ordovician.

Franziska Franeck. Natural History Museum, University of Oslo, Post Office Box 1172, Blindern, N-0318, Oslo, Norway. E-mail: [email protected] Lee Hsiang Liow. Natural History Museum, University of Oslo, Post Office Box 1172, Blindern, N-0318, Oslo, Norway; Centre for Ecological and Evolutionary Synthesis, Department of Biosciences, University of Oslo, Post Office Box 1066, Blindern, N-0316, Oslo, Norway. E-mail: [email protected]

Accepted: 21 January 2019 First published online: 11 March 2019 Data available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.nk218ht

fi Introduction to group them into two classes. The rst class of hypotheses is climate related. Specifically, a The numerical increase of marine orders, mid-Ordovician cooling is thought to have con- families, and genera in the Ordovician was so tributed to a more favorable global climate dramatic (Sepkoski et al. 1981; Miller 1997a,b, regime that in turn promoted diversification 2012; Sepkoski 1997; Harper 2006; Bassett (Trotter et al. 2008; Rasmussen et al. 2016). et al. 2007; Alroy et al. 2008) that it has been The second class of hypotheses involves nutri- termed the great Ordovician biodiversification ent availability. For instance, periods of event (GOBE; Webby et al. 2004; Servais and increased tectonics (Miller and Mao 1995) Harper 2018). This increase is thought to be and/or tectonically induced volcanic activity especially rapid around the Dapingian/Darri- (Botting 2002) are hypothesized to have led to wilian boundary (Servais et al. 2009; Hints more sedimentary and nutritional input into et al. 2010; Rasmussen et al. 2016; Trubovitz the oceans and habitat fractioning, both of and Stigall 2016), but factors influencing this which may have enhanced the diversification increase remain obscure. of taxa (Miller and Mao 1995). The establish- While the hypotheses for what these factors ment of nutrient-rich upwelling zones has might be are many and varied, it is possible been documented from Laurentia during the

© 2019 The Paleontological Society. All rights reserved. This is an Open Access article, distributed under the terms of the DownloadedCreative from https://www.cambridge.org/core Commons Attribution licence. IP address: (http://creativecommons.org/licenses/by/4.0/), 170.106.34.90, on 26 Sep 2021 at 18:44:57, subject to the which Cambridge permits Core terms unrestricted of use, available re- at https://www.cambridge.org/core/termsuse, distribution, and reproduction. https://doi.org/10.1017/pab.2019.4 in any medium, provided the original work is properly cited. 0094-8373/19 222 FRANZISKA FRANECK AND LEE HSIANG LIOW

Middle Ordovician (Pope and Steffen 2003). changes in sea levels lead to changes in the These upwelling zones may have served as size of habitable areas (Holland and Christie new ecospace (Rasmussen et al. 2016), and 2013). Here, we explore whether greater envir- their subsequent occupation by both migrants onmental variability in onshore areas is asso- and taxa that evolved in situ could have con- ciated with higher genus origination and tributed to an increase in taxon diversity. extinction rates compared with offshore areas Some of the mechanisms that were suggested during the Ordovician. to have an influence on the GOBE, such as tec- We present Ordovician genus origination tonic activity and volcanism, likely acted at a and extinction rates based on capture–recap- regional rather than the global scale, and such ture models for the revised global Ordovician regional changes are unlikely to be simultan- stages (Bergström et al. 2009; Harper and Ser- eous (Miller 1997a,b, 2004; Zhan and Harper vais 2013; Lindskog et al. 2017) both on a global 2006). Therefore the influence of the different scale, for the paleocontinents Baltica and Laur- mechanisms should be examined separately entia, and for onshore and offshore areas. for the different paleocontinents (Miller 1997a; Miller and Mao 1998) and paleoenvironments Methods and Data (e.g., Miller and Mao 1995; Miller and Connolly 2001; Novack-Gottshall and Miller 2003). Data.—We downloaded data from the Paleo- Continents drifted and experienced different biology Database (PBDB) on 26 January 2018 spatial environments over the Ordovician. (https://paleobiodb.org/data1.2/occs/list.csv? Whereas the center of the paleocontinent Laur- taxon_reso=genus&interval=Cambrian,Aeronian entia was situated close to the equator through- &show=class,acconly,ecospace,coll,coords,loc, out the Ordovician, Baltica rotated northward, paleoloc,lithext,geo,refattr). These data consist starting in the southern mid-latitudes and end- of 130,367 occurrences of taxa identified to at ing south of the equator by the Late Ordovician least the genus level that span the Cambrian to (Cocks and Torsvik 2006; Torsvik and Cocks the Aeronian (second stage in the Silurian) and 2016a). These two continents can act as model their metadata. Only accepted genus names systems for a relatively stable continent (Laur- were included in our data analyses (73,735 entia) versus a continent on which the physical occurrences). Although our analyses are focused environmental is relatively unstable due to con- on the Ordovician, we included Cambrian and tinental movement (Baltica) during the time in Silurian data to ameliorate edge effects (see sub- question. In this paper, we explore whether section on diversification rates). greater environmental change on Baltica is The temporal resolution associated with the associated with greater taxon turnover but data we retained from the PBDB download lower genus richness compared with Laurentia. range from 0.4 to 66.2 Myr. Considering only Different paleoenvironments may also influ- Ordovician time intervals, the resolution is ence diversification rates. For instance, mor- from 1.7 to 50.8 Myr (median: 7.8 Myr; mean: phological novelties, as represented by 8.6 Myr). To increase the temporal resolution ordinal-level originations, arose preferentially of our analyses, we only included PBDB data in onshore rather than offshore areas (Jablonski that were assigned to time bins smaller than et al. 1983; Jablonski 2005). Based on his work 12 Myr (see Supplementary Material text and on ordinal-level originations, Jablonski Supplementary Fig. S1). For occurrences (Jablonski et al. 1983; Jablonski 2005) predicted reported in PBDB with regional Ordovician that genus originations would preferentially stages, we randomly assigned point estimates occur where their higher-level taxa were using a uniform distribution between the already established. If this is true, we expect reported minimum and maximum ages and that relative origination dynamics might be then assigned those estimates to global Ordovi- similar for onshore and offshore areas, all cian stages (median: 7.7 Myr; mean: 5.9 Myr). things being otherwise equal. However, off- This was to ensure comparability to standard, shore areas are more shielded from environ- global treatments of Ordovician data. We pre- mental changes than onshore areas, where sent only one iteration of parameter estimates

Downloaded from https://www.cambridge.org/core. IP address: 170.106.34.90, on 26 Sep 2021 at 18:44:57, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/pab.2019.4 DIVERSIFICATION DYNAMICS OF THE GOBE 223

in our main figures. But because random question. Similarly, seniority probability γ assignments to time intervals vary, we also pre- describes the probability that if a genus was sent averaged parameter estimates from 100 extant in time i+1, it was already extant in iterations of such random assignments to the time i. In this case, 1 − γ is the origination prob- global data set or subsets for each of our ana- ability between the two time intervals. Diversi- lyses in the Supplementary Material. fication λ describes how much a group of taxa is We created a non-observation/observation growing or declining from one time interval to fi λ matrix from the downloaded PBDB data. In the next, where the net diversi cation rate is i this matrix, each row of data is a string of obser- − 1. If the number of genera increases from one vations (1) or non-observations (0) of a genus time interval (i) to the next (i+1), the net diver- within sequential global stages. We estimated sification rate will be positive, while if the num- extinction, origination, and sampling rates by ber of genera is decreasing, it will be negative. applying a capture–recapture modeling frame- A value of zero indicates no net increase or work to these data. Although there are many decrease in the number of genera. Given that different models within the capture–recapture the Pradel seniority model has been previously modeling framework (Nichols and Pollock detailed for paleontological data sets (Connolly 1983; Liow and Nichols 2010), we follow previ- and Miller 2001a,b, 2002; Liow and Nichols ous paleontological applications (Connolly and 2010; Liow et al. 2015), we refer readers to Miller 2001b; Liow et al. 2015; Kröger 2017)in those publications for further details. It is suffi- using the Pradel seniority model (Pradel 1996) cient to note that a logit link is used and that the for estimating diversification rates. In addition, parameters are estimated in a maximum- we use the POPAN model (Schwarz and likelihood framework. Arnason 1996) for estimating genus richness. As already mentioned, the basic parameters We note that up to 65% of our global genus of the Pradel seniority model are survival, seni- data set is made up of arthropods, brachiopods, ority, and sampling probabilities. These basic and mollusks. The remaining 35% of the parameters can be specified in different ways. phyla in the data set are bryozoans, chordates, For example, we can assume that they are con- cnidarians, echinoderms, hemichordates, and stant through time (time-constant model). poriferans. Using origination probability as an example, Diversification Rates.—The Pradel seniority this means that time interval i has the same ori- model (Pradel 1996) was developed for estimat- gination probability as time interval i+1, and ing population parameters, namely survival so on, such that there is only one estimate for probability w, seniority probability γ, sampling the whole of the Ordovician. Because event probability p, and population growth λ. probabilities increase as the sampling intervals Because our data represent genera rather than increase, we specify wt or γt, with t being the individuals, these parameters reflect, respect- duration between two sampling occasions ively, genus survival probability (the comple- such that origination and extinction probabil- ment of which is genus extinction ities can be interpreted in this time-constant probability), genus “seniority” probability (the model. Another model we could use is a fully complement of which is genus origination time-varying one, meaning that estimates probability), sampling probability, and diversi- (e.g., for origination) are allowed to be different fication rates. Very briefly, we assume that gen- for every time interval. Here, to allow compar- era are sampled within time intervals, i.We isons of our estimates across these stages, we number these from older (i) to younger geo- transform the estimated probabilities into logical intervals (i+1). In this framework, sur- rates by using a Poisson model. Specifically, w Φ − − −w vival ( ) is the probability that a genus that is i = log[1 (1 )i]/Ti for the extinction rate Φ Γ − − −γ extant in a given time interval continues to be i, i = log[1 (1 )i]/Ti for the origination Γ − − − extant in the next time interval. Extinction rate i, and Pi = log[1 (1 p)i]/Ti for the − w probability (1 ) is hence the probability of a sampling rate Pi (Liow et al. 2015), with Ti genus going extinct in the next time interval, being the duration of one time interval. Add- given that it was extant in the time interval in itionally, we can add covariates to the model,

fi such as af liation with Laurentia or Baltica, in and pentmax−1. Thus, we include additional order to estimate separate parameters for the time bins before and after the Ordovician, as two groups simultaneously. We specified and done for the Pradel seniority model. examined many models (see “Results”) but pre- Laurentia and Baltica.—To assign the obser- sent only a selection of these in our main text vations of genera documented in the PBDB to (see Supplementary Material for additional paleocontinents, we looked up their geoplate results). code from Müller et al. (2008), which is In a fully time-varying Pradel seniority provided in the PBDB, and then assigned model, survival and sampling in the last time these to either Baltica or Laurentia using w interval (given by pi i−1, with i in this case Cocks and Torsvik (2004) and Torsvik and being the last time interval) and seniority and Cocks (2016a,b) for guidance. Note that in sampling in the first time interval are not separ- some cases, geoplate codes were not assigned ately identifiable (Connolly and Miller 2001b). in the PBDB. In those cases, we used country To obtain estimates for survival, seniority, (cc) or state, encoding modern positions, for and sampling probabilities for all of the Ordo- assignments to paleocontinents in PBDB. We vician, we include two Cambrian time bins provide our assignments in the Supplementary before and two Silurian after the first and last Material. Some genera were observed as Ordovician stages, respectively. In that way “endemic” to Baltica (194 genera) or Laurentia we avoid the described boundary effects. (731 genera), while others were described Note that for the time-constant model, we from both paleocontinents (480 genera). We only use occurrence data from the Ordovician, restricted our analyses to those genera that because there are no boundary effects and were observed only on either one of the paleo- because we are only interested in comparing continents in question. “Endemic” is in quota- Ordovician probabilities. tion marks, because we assume that genera Genus Richness.—We estimated genus rich- are found only where they are observed, and ness using the POPAN model (Schwarz and we do not explicitly model migration. Arnason 1996). In contrast to the closed- We note that 33% of the Baltic genera are population approach of the Pradel seniority arthropods, while 18% are brachiopods and model, the POPAN model is an open- 16% are mollusks. The remaining 33% of the population approach (Schwarz and Arnason Baltic genera are bryozoans, echinoderms, cni- 1996). This means that individuals (genera in darians, chordates, annelids, hemichordates, our case) can enter the observed population and others. Of the Laurentian genera, 25% are (global or regional community in our case) arthropods, 21% are mollusks, and 19% are from a superpopulation. In our context, the echinoderms. The remaining 35% of Laurentian superpopulation (N ) is the number of genera genera are bryozoans, brachiopods, poriferans, available to be sampled that may or may not cnidarians, chordates, and others. More be sampled throughout the whole study per- detailed information on the representation of = i−1 iod, where N 0 Bi (Schwarz and Arnason different phyla in the data subsets can be – 1996). Here, Bi is the number of new genera ori- found in Supplementary Tables S6 S8 in the ginating between two time intervals (number Supplementary Material. of births in the original description of Onshore–Offshore.—For our comparison of POPAN). It is estimated by Bi = penti−1N, onshore and offshore areas, we assign genera where pent is the probability of entering using information on depositional environ- the pool of genera. The genus richness in ments given in the PBDB (see Supplementary w each time interval is then Ni = Ni−1 i−1 + Bi Material for details). Roughly speaking, we w (Schwarz and Arnason 1996), where i−1 is consider onshore environments as those the survival probability into the time interval above fair-weather wave base in high-energy in focus. Similar to the edge effect described environments and offshore environments as w above, i−1 cannot be estimated independently those below storm wave base representing low- in a fully time-varying POPAN model. This is energy environments. We only consider genera fi also true for the parameters N1, Nmax, pent1, that have their rst occurrence record in either

onshore (532 genera) or offshore (396 genera) of genera about doubled from 900 in the Dapin- areas, regardless of the locations of their subse- gian to 1900 in the Darriwilian. After the quent observations. This means that extinction increase in the Darriwilian, genus richness rates that are estimated with these data are stays relatively high until the Hirnantian extinc- not representative for the actual extinction tion. There is an average increasing trend in that happens in onshore and offshore areas, sampling rates throughout the Ordovician respectively, because taxa may have migrated (Fig. 1C). and became extinct somewhere else. To esti- Origination probabilities for Baltica are 0.190 mate extinction rates for onshore and offshore (95% CI 0.155, 0.230) per million years for the areas, we considered genera that have their whole Ordovician, while Laurentian origin- last occurrence records in either onshore (537 ation probabilities are 0.088 (95% CI 0.079, genera) or offshore (411 genera) areas, regard- 0.097), given a Pradel seniority model that less of the locations of their prior observations. imposes constant origination, extinction, and In doing so, we assume that these first and last sampling probabilities through time. Extinction paleoenvironments of observations are the probabilities, estimated with the same time- same as those where they actually originated constant model, are 0.169 (95% CI 0.134, or went extinct. 0.211) for Baltica and 0.069 (95% CI 0.061, The analyses were done with the program 0.080) for Laurentia. Net diversification prob- MARK (White 2016) via the RMark interface abilities are 0.026 (95% CI 0.014, 0.038) for for R (Laake 2013). Baltica and 0.020 (95% CI 0.014, 0.025) for Laurentia. Baltica shows greater interval-to-interval ori- Results gination rates than Laurentia throughout the Global origination rates are similar to extinc- Ordovician (Fig. 2A). At the Tremadocian/ tion rates between the Tremadocian and Dapin- Floian and Dapingian/Darriwilian boundar- gian (Fig. 1A). At the Dapingian/Darriwilian ies, origination rates are at their highest on Bal- boundary, the global origination rate is much tica. On Laurentia, origination rates are at their higher than the extinction rate, which also highest at the Cambrian/Tremadocian and dropped perceptibly compared with the earlier Dapingian/Darriwilian boundaries. Note that Ordovician. Extinction and origination rates variation in estimates and their 95% confidence become similar again before the Hirnantian intervals is much greater until the Dapingian/ mass extinction (Fig. 1A). Darriwilian boundary (see also Supplementary The resulting net diversification is negative Fig. S7). However, after the Dapingian/Darri- between the Tremadocian and Dapingian, cor- wilian boundary, Baltic origination rates are responding to the extinction rates being slightly consistently higher than Laurentian ones. greater than origination rates during these time From the Tremadocian/Floian boundary intervals (Fig. 1B). Between the Dapingian and and thereafter, extinction rates are higher on Darriwilian, net diversification dramatically Baltica than on Laurentia (Fig. 2B). On both increases, with the number of genera doubled. paleocontinents, extinction rates decrease on Net diversification is lower again thereafter average from the Tremadocian/Floian bound- (i.e., close to zero), before it becomes negative, ary until the Darriwilian/Sandbian boundary. driven by the high extinction rate during Thereafter they show an increasing trend. the Hirnantian mass extinction. Estimates of Note that the 95% confidence intervals for genus richness largely correspond to the extinction rates are largely overlapping for changes in net diversification changes the two paleocontinents until the Dapingian/ (Fig. 1B), despite different assumptions of Darriwilian boundary (cf. Supplementary “population closure” used to estimate diversifi- Fig. S7). Thereafter, Baltic extinction rates are cation and richness (see “Methods”). For consistently higher than Laurentian extinction example, net diversification rate between the rates. Dapingian and Darriwilian is about 1; hence As a result of the extinction and origination it follows that the point estimate of the number rates (Fig. 2A,B), net diversification rates are

FIGURE 1. Global genus diversification and genus diversity dynamics. A, Origination (solid line and solid circles, O) and extinction (dotted line and open circles, E) rates as events per million years (Myr). B, Net diversification rate (solid line and triangles) and genus richness (dotted line and rectangles). Horizontal gray line represents zero net diversification. C, Sam- pling events per million years (Myr). Vertical lines are 95% confidence intervals. Stage abbreviations: Tr, Tremadocian; Fl, Floian; Dp, Dapingian; Dw, Darriwilian; Sa, Sandbian; Ka, Katian; Hi, Hirnantian.

greater on Baltica than on Laurentia at the Tre- same time, although the trajectories of both madocian/Floian boundary (Fig. 2C). At the paleocontinents are very similar. The peak in Floian/Dapingian boundary, net diversifica- net diversification (cf. Fig. 2) is expressed as a tion rates are slightly greater on Laurentia fivefold increase of genus richness on Baltica than on Baltica. Thereafter, net diversification and an almost threefold increase on Laurentia rates are slightly greater on Baltica than on between the Dapingian and Darriwilian. This Laurentia, until the Katian/Hirnantian bound- results in an increase from about 13 to 64 genera ary, although 95% confidence intervals are on Baltica, and from about 99 to 286 genera overlapping between the two paleocontinents on Laurentia for the higher taxa we examined. throughout. Although the first substantial increase in Sampling rates increase on average from the genus richness happened at the Dapingian/ Floian to the Hirnantian for Laurentia Darriwilian boundary, net diversification rates (Fig. 2D). Sampling rates for Baltica can only close to or slightly above zero until the be estimated from the Darriwilian onward, Katian/Hirnantian boundary (Fig. 2C) led to during which sampling rates are lower on an Ordovician peak in genus richness during Baltica compared with Laurentia. the Katian (Fig. 3). Baltica had relatively low genus richness in Origination and extinction rates in onshore the Early and early Middle Ordovician (Trema- and offshore areas are barely distinguishable docian to Dapingian) (Fig. 3). Laurentia had throughout the Ordovician. Onshore origin- greater genus richness than Baltica during the ation probabilities are 0.069 (95% CI 0.060,

FIGURE 2. Genus diversification dynamics on Baltica and Laurentia. A, Origination rates as events per million years (Myr). B, Extinction rates as events per million years (Myr). C, Net diversification rates where the horizontal line represents zero net diversification. D, Sampling rates as events per million years (Myr). Vertical lines represent 95% confidence intervals in A to D. Baltic rates (B) are indicated by solid lines and circles, while Laurentian rates (L) are indicated by dashed lines and triangles. Stage abbreviations as in Fig. 1. Estimated probabilities or confidence intervals thereof that are one or zero are indications of poorly constrained estimates and are hence removed from the plots. Net diversification rates with confidence intervals spanning more than 10 are also removed for the same reason.

0.079) per million years for the whole Ordovi- cian, while offshore origination probabilities are 0.060 (95% CI 0.052, 0.070), given a Pradel seniority model that imposes constant origination, extinction, and sampling probabilities through time. Extinction probabilities for the whole Ordovician, using the reversed data set but the same model, are 0.070 (95% CI 0.058, 0.085) for onshore areas and 0.085 (95% CI 0.069, 0.104) for offshore areas. Interval-to-interval origination rates in onshore and offshore areas are very similar throughout most of the Ordovician, given a FIGURE 3. Genus diversity dynamics on Baltica (B) and time-varying Pradel seniority model with sep- Laurentia (L). Estimated genus richness on Laurentia arate estimates for onshore and offshore taxa (dashed line, triangles) and Baltica (solid line, circles) based on the POPAN model. Stage abbreviations as in (Fig. 4). After the Cambrian/Tremadocian Fig. 1. Vertical lines represent 95% conﬁdence intervals. boundary, onshore rates, which start higher

our hypothesis, see Supplementary Material (Supplementary Figs. S9 and S10). In this study we are specifically interested in changes of diversification rates on different paleocontinents or onshore/offshore areas through time. Therefore, we choose to present results from the fully time-varying model for all parameters and covariates. For completeness, we tested the fit of different models with different parameter specifications through time by comparing their Akaike information criterion values. Results of these model com- parisons can be found in the Supplementary Material (Supplementary Tables S1–S4).

Discussion Global Ordovician Dynamics.—We confirmed, by explicitly modeling incomplete sampling, that the increase of genus richness during the Ordovician was very rapid around the Dapin- gian/Darriwilian boundary, as suggested by FIGURE 4. Genus origination and extinction rates in many authors, including Servais et al. (2009), onshore and offshore areas. A, Origination rates as events per million years (Myr), based on first observation in either Hints et al. (2010), Harper et al. (2013), Rasmus- onshore or offshore areas. B, Extinction rates as events per sen et al. (2016), and Trubovitz and Stigall million years (Myr) based on last observation in either (2016). However, this increase could have onshore or offshore areas. Vertical lines represent 95% confidence intervals in A and B. Onshore rates (on) are indi- been driven by a lowered extinction rate and/ cated by solid lines and circles while offshore rates (off) or a higher origination rate. Here, we show are indicated by dashed lines and triangles. Stage abbrevia- for the first time, that global origination rates tions as in Fig. 1. Estimated probabilities or confidence intervals thereof that are one or zero are indications of for the Ordovician peaked at the Dapingian/ poorly constrained estimates and are hence removed from Darriwilian boundary, right before the spike fi fi the plots. Net diversi cation rates with con dence intervals in genus richness during the Darriwilian spanning more than 10 are also removed for the same reason. (Fig. 1). In addition, the extinction rate at the same boundary is slightly lower than in the previous boundary (Floian/Dapingian bound- than offshore rates, converge. At the Dapin- ary). Net diversification rates were in fact close gian/Darriwilian boundary, origination rates to or lower than zero before the Dapingian/ are high in both paleoenvironments, reflecting Darriwilian boundary, after which they the global dynamics at that time. rocketed up with a doubling of genera (from Until the Floian/Dapingian boundary, c. 800 to c. 1900 genera; Fig. 1) during the Dar- extinction rates are higher in onshore areas riwilian. This doubling of genera is greater than than in offshore areas (Fig. 4B). At the Dapin- what was inferred recently by Kröger (2017) gian/Darriwilian boundary, extinction rates using a similar approach. Kröger estimated a from onshore and offshore areas are similar, 1.4 times increase in genus richness over the with greatly overlapping confidence intervals. same time period (see Kröger 2017), although Thereafter, until the Katian/Hirnantian bound- his genus richness estimate for the Darriwilian ary, extinction rates are higher in offshore than (ca. 2000) is similar to ours (ca. 1900). In add- in onshore areas. ition, Kröger’s maximum genus richness for For diversification and sampling rates, based Ordovician stages stands at ca. 2200 genera on first and last observations in either onshore during the Katian, in contrast to ours, which or offshore areas, which are not central for occurs at the Darriwilian. These differences,

despite our using the same model for genus The general diversification trends estimated in richness inference, might be due to differential Connolly and Miller may also differ from ours temporal assignments of genus records from due the broader set of marine invertebrates the PBDB via RNames in the Kröger study. used in our estimation. For completeness, we RNames (Kröger and Lintulaakso 2017)isa divided our data set into subsets of those taxa dynamic interface that correlates selected strati- Connolly and Miller (2001a) studied (see graphic opinions with the purpose of increas- Supplementary Figs. S12, S15–S17). We find ing stratigraphic resolution. Dividing stages that the dynamics of these taxa broadly con- into stage slices reduces the number of observa- form to the global patterns we have estimated. tions per time bin, possibly depleting some However, others have shown that graptolites stage slices of data, so confidence intervals of (Cooper et al. 2014) and echinoderms (Sprinkle the estimates would subsequently increase, and Guensburg 1995, 2004), unlike the taxa we especially for our paleocontinental and paleo- investigated (arthropods, brachiopods, and environmental analyses. The increases in confi- mollusks; see Supplementary Material), show dence intervals would be even more severe if an increase in genus richness during the Early observations are restricted to specific paleocon- Ordovician. Given that the model used in tinents or bathymetric groups. In addition, the Connolly and Miller (2001b) and this study rules by which stages are subdivided into explicitly accounts for sampling probabilities, stage slices in RNames are currently not well- we do not expect the addition of more observa- documented enough for our use. For these tions (of correctly identified and dated samples) reasons, we estimated diversification rates on to change the estimates of mean extinction a coarser time resolution, but with better and origination considerably, but to increase understood data for the benefit of more confi- the sampling probabilities while tightening dent estimates. There is some concern that the confidence intervals for extinction and bias is introduced by assigning fossil occur- origination. rences to global stages when the occurrences As already mentioned, many hypotheses were originally associated with regional stages. have been raised for why diversification rates However, our sensitivity analyses show that were so high during the GOBE, some of reducing the data set to only fossil occurrences which likely acted more globally and others assigned to global stages does not change our more regionally. While our global estimates qualitative inferences (Supplementary Fig. S6 integrate over the processes that occur at local in the Supplementary Material). to global levels, details of regional dynamics In their pioneering global study, Connolly give us clues as to what types of processes and Miller (2001b) simultaneously estimated might be operating. In other words, if the origination, extinction, and sampling probabil- GOBE in different regions is driven by different ities for the Ordovician. They found that both factors, we would expect diversification origination and extinction probabilities have dynamic patterns to be different. For instance, higher values approximately around the Darri- if increased sedimentary input influenced wilian, although they found no exceptional diversification rates, the rates of taxa closest to diversification peak (Connolly and Miller volcanically and tectonically active areas may 2001b), in contrast to what we found. While experience more changes than those in the Connolly and Miller (2001b) used the same oceans farther away, all other things being model we did, they estimated origination and equal. To explore the contrast between global extinction probabilities between time intervals and regional controls of Ordovician diversifica- of equal length, because probabilities of events tion, we examine the dynamics on Laurentia over unequal time intervals cannot be directly and Baltica and in onshore and offshore areas compared. Rather than forcing these data to and discuss these in relation to our global conform to geologically unnatural but equal estimates. time intervals, we used unequal but geologic- Contrasting Laurentia and Baltica.—It is strik- ally natural Ordovician stages, but we con- ing that the peaks in origination rates happened verted probabilities into rates (see “Methods”). on both Baltica and Laurentia between the

Dapingian and the Darriwilian (Fig. 2), a pro- to specific organismal groups. For instance, cess that led to the largest increase in genus Trubovitz and Stigall (2016) found a peak in richness over the Ordovician at the Darriwilian brachiopod species richness on both Baltica for both continents. Prior to this peak, origin- and Laurentia in the middle Darriwilian by ation and extinction rates appear to be uncoor- using range-through data compiled by Ras- dinated on the two paleocontinents (see also mussen et al. (2007). Similarly, Paris et al. Supplementary Fig. S7). After the Dapingian/ (2004) recorded the onset of the GOBE in Darriwilian boundary, changes in diversifica- chitinozoans at the Dapingian/Darriwilian tion dynamics appear to be more coordinated. boundary on both Baltica and Laurentia, Based on this observation, we suggest that we based on range-through data. It is noteworthy can subdivide Ordovician diversification into that their actual peak of Laurentian chitinozo- three different phases. In the first phase (the ans was found in the early Late Ordovician, Early Ordovician: Cambrian/Tremadocian while the Baltic peak was detected spanning and Tremadocian/Floian boundaries), the vari- from the Darriwilian to the early Sandbian ation in diversification rates seems relatively (Paris et al. 2004). These findings are similar to uncoordinated on Baltica and Laurentia. In our finding that peak genus richness occurred the second phase (the Middle Ordovician: in the Katian Laurentia and Baltica (Fig. 3). Floian/Dapingian and Dapingian/Darriwilian While the diversification dynamics on Laur- boundaries), origination rates peak dramatic- entia and Baltica are broadly similar after the ally on both paleocontinents. In the last phase Dapingian/Darriwilian boundary, there are (the Late Ordovician: Darriwilian/Sandbian also differences between the two paleoconti- to Katian/Hirnantian boundaries), diversifica- nents. At the Tremadocian/Floian boundary, tion rates seem to change in concert on the there appear to be peaks in both origination two paleocontinents. The three phases hint at and extinction rates on Baltica that are not regional effects on diversification being expressed on Laurentia. Given the size of the stronger in the Early Ordovician, and global 95% confidence intervals (see also Supplemen- effects being stronger in the Middle to Late tary Fig. S7), we refrain from overinterpreting Ordovician. We note in passing that our these observations. However, these peaks onshore and offshore data subsets appear also may indicate changes on Baltica before the to be more uncoordinated before the Dapin- Dapingian/Darriwilian peak that enhanced gian/Darriwilian boundary than after that, increased turnover, although this did not result just like the dynamics we estimated for the in an increase in genus richness. two paleocontinents. The abovementioned regional effects could In the Early Ordovician, Laurentia was stably have been dominating factors in the early situated close to the equator, while Baltica was Ordovician, while globally acting changes rotating from the southern midlatitudes toward could have contributed to a coordinated lower latitudes (Torsvik and Cocks 2016a), increase of origination rates on Baltica and moving through different climatic zones. Laurentia during the GOBE (i.e., starting in Active volcanism occurred during the whole the Dapingian and continuing into the Darriwi- Ordovician (Barnes 2004), accompanied by lian). One such global change suggested as a great siliciclastic input in adjacent areas (Ser- driver of the GOBE is the proposed cooling vais et al. 2009), especially on the eastern side that began during the Middle Ordovician (Trot- of Laurentia, where the Taconic orogeny had ter et al. 2008; Rasmussen et al. 2016) and was started during the Middle Ordovician (Holland sustained until the Late Ordovician (e.g., Saltz- and Patzkowsky 1996; Novack-Gottshall and man and Young 2005; Trotter et al. 2008). The Miller 2003). This greater siliciclastic input suggested timing of the onset of cooling fits purportedly led to a greater genus richness on broadly with the increase in origination rates Laurentia (Miller 1997a). The increase in that we have found in our study. Frequent genus richness from the Dapingian to the asteroid impacts during the Early and Middle Darriwilian common to Laurentia and Baltica Ordovician represent another worldwide con- corroborates previous observations restricted dition that would have been experienced by

the Ordovician seas (Schmitz et al. 2001), and increases substantially between Darriwilian these extraterrestrial inputs have also been and Katian, despite the Pradel-estimated net linked to GOBE (Schmitz et al. 2008). This diversification rates on Laurentia only being asteroid-driven GOBE hypothesis has recently slightly above zero. been refuted on the basis of refined timings of Contrasting Onshore–Offshore Environments.— these impacts using new isotopic data (Linds- In addition to examining paleocontinental kog et al. 2017), but we argue that we cannot differences, we also explored origination completely reject this hypothesis before we dynamics in two contrasting environments have temporal estimates of diversification that (onshore and offshore) in order to understand are also more highly resolved. what may influence diversification rates. In Ecological theory and contemporary empir- addition to differing energy levels, sea-level ical studies have shown that larger terrestrial changes may also more greatly affect onshore areas can harbor more species than smaller areas than offshore areas (Holland and Christie areas (e.g., MacArthur and Wilson 1967; 2013). Rosenzweig 1995). However, this relationship Jablonski suggested that origination at the is less clear for the marine realm (Roy et al. genus level should occur mostly where the 1998; Valentine 2009), in part due to fluid higher-level taxa in question were already boundaries. Despite uncertainties, we were established (Jablonski et al. 1983; Jablonski still curious as to whether we could detect a 2005), while Tomašových et al. (2014) found greater total genus richness on Laurentia than evidence, using data from the Eocene and Plio- Baltica (as found by Miller 1997a), given that Pleistocene, that genus origination and extinc- circumstantial evidence indicates that Lauren- tion rates in onshore areas are greater than tia was much larger than Baltica (Torsvik and those in offshore areas. We note that Ordovi- Cocks 2016a). Here, we assume, as in many cian origination and extinction rates are only paleontological studies (e.g., Sepkoski 1991; minutely different between on- and offshore Finnegan and Droser 2005), that we can use environments (Fig. 4), results that are more con- genus richness as a proxy for species richness. sistent with Jablonski’s hypotheses than with The estimated genus richness for the Ordovi- Tomašových’s. Analyses based on model selec- cian differs quite substantially between the tion giving extra support for the similarity in two paleocontinents (Laurentia, 646 genera origination rates between onshore and offshore [95% CI 627, 683]; Baltica, 206 genera [95% CI areas are presented in the Supplementary 185, 294]), supporting our naive expectation. Material. In addition, there is also temporal variation In comparing onshore and offshore genus that is not coordinated through time (Fig. 3). origination rates, we had to assume that any We restricted our paleocontinental analyses unobserved occurrences preceding the first to genera that were observed on either of the record of the genus in question in a given envir- paleocontinents but not both. In doing so in onment are in the same environment as the first our analysis framework, we assumed that the record. Given the incomplete temporal sam- genera in question actually originated on their pling of any genus, we realize this may not be assigned paleocontinents and that there are a robust assumption, but we lack the tools to no unobserved occurrences on other paleocon- deal with this uncertainty at this time. How- tinents prior to our first observations. We are ever, given that we discarded genera with mul- aware that this assumption might be violated tiple paleoenvironmental associations in their for some genera, but the Pradel seniority first time interval of observation in the PBDB, model does not consider observed or unob- we believe that “rogue” genera that occur in dif- served migration. In addition, we want to ferent environments before their first observa- emphasize that while the Pradel seniority tion are rare. An additional caveat is that model assumes a closed population, the environmental classifications in the PBDB are POPAN model assumes an open population. not clearly defined, such that anyone entering This may explain why the POPAN-estimated data into the PBDB may have to make a subject- genus richness on Laurentia continuously ive decision as to which environmental

classification to assign to the fossil in question. K. L. Voje, T. Reitan, P. Smits, J. P. Nystuen, S. As a result of uncertainty, many environmental Finnegan, and B.Kröger for discussions. We assignments end up in the “indet.” (indeter- thank J. Crampton, A. Miller, and two anonym- minate) category. The onshore and offshore ous reviewers for their constructive feedback data set we used is quite a lot smaller than and comments that helped improve this article. what was potentially available (see Supple- We also want to thank all those who contribu- mentary Material). However, we have no rea- ted primary literature and data to the PBDB son to suspect that data points for which and the members of the PBDB for providing environment is marked “indet.” in the PBDB data for our analysis. This article is a contribu- are biased toward any of the two environments tion to IGCP Project 653: “The Onset of the in which we are interested. We are also aware Great Ordovician Biodiversification Event.” that our division of onshore and offshore envir- This is PBDB publication number 331. onments is more conservative than others published (e.g., Bottjer and Jablonski 1988) Literature Cited and that a different subset of data might Alroy, J., M. Aberhan, D. J. Bottjer, M. Foote, F. T. Fürsich, P. suggest different results. J. Harries, A. J. W. Hendy, S. M. Holland, L. C. Ivany, W. Kiessling, M. A. Kosnik, C. R. Marshall, A. J. McGowan, A. I. Miller, T. D. Olszewski, M. E. Patzkowsky, S. E. Peters, Conclusions L. Villier, P. J. Wagner, N. Bonuso, P. S. Borkow, B. Brenneis, M. E. Clapham, L. M. Fall, C. A. Ferguson, V. L. Hanson, A. The global increase in genus richness at the Z. Krug, K. M. Layou, E. H. Leckey, S. Nurnberg, C. M. Powers, Dapingian/Darriwilian boundary was the result J. A. Sessa, C. Simpson, A. Tomašových, and C. C. Visaggi. 2008. Phanerozoic trends in the global diversity of marine inver- of a dramatic increase in origination rates, while tebrates. Science 321:97–100. extinction rates stayed at relatively low levels. Barnes, C. R. 2004. Was there an Ordovician superplume event? This global increase in origination rates is Pp. 77–80 in Webby et al. 2004. Bassett, D., K. G. MacLeod, J. F. Miller, and R. L. Ethington. 2007. mirrored on Laurentia and Baltica, as well as in Oxygen isotopic composition of biogenic phosphate and the onshore and offshore areas, pointing toward a temperature of Early Ordovician seawater. Palaios 22:98–103. global mechanism behind the peak in origin- Bergström, S. M., X. Chen, J. C. Gutiérrez-Marco, and A. Dronov. 2009. The new chronostratigraphic classification of the Ordovi- ation rates at the Dapingian/Darriwilian bound- cian System and its relations to major regional series and stages ary, which is likely the greatest contributor to the and to δ13C chemostratigraphy. Lethaia 42:97–107. GOBE. Based on the patterns of paleocontinental Botting, J. P. 2002. The role of pyroclastic volcanism in Ordovician fi diversification. In J. A. Crame and A. W. Owen, eds. Palaeobio- and paleoenvironmental diversi cation dynam- geography and biodiversity change: the Ordovician and Meso- ics, it appears that regionally varying factors zoic–Cenozoic radiations. Geological Society of London Special were prominent during the Early Ordovician, Publication 194:99–113. Bottjer, D. J., and D. Jablonski. 1988. Paleoenvironmental patterns while globally acting controlling factors appear in the evolution of post-Paleozoic benthic marine invertebrates. to have been dominant during the Middle Palaios 3:540–560. Ordovician and onward. All else being equal, Cocks, L. R. M., and T. H. Torsvik. 2004. Major terranes in the Ordovician. Pp. 61–67 in Webby et al. 2004. our results suggest that origination probabilities ——. 2006: European geography in a global context from the are slightly greater in areas exposed to greater Vendian to the end of the Palaeozoic. In D. G. Gee and environmental change (i.e., greater disturbance R. A. Stephenson, eds. European lithosphere dynamics. Geological Society of London Memoir 32:83–95. through sea-level changes or continental Connolly, S. R., and A. I. Miller. 2001a. Global Ordovician faunal movement). transitions in the marine benthos: proximate causes. Paleobiology 27:779–795. ——. 2001b. Joint estimation of sampling and turnover rates from Acknowledgments fossil databases: capture-mark-recapture methods revisited. Paleobiology 27:751–767. This work was funded by the NFR project ——. 2002. Global Ordovician faunal transitions in the marine 235073/F20 (principal investigator: L.H.L.). benthos: ultimate causes. Paleobiology 28:26–40. š Cooper, R. A., P. M. Sadler, A. Munnecke, and J. S. Crampton. We thank A. Toma ových for helpful com- 2014. Graptoloid evolutionary rates track Ordovician–Silurian ments on the division of onshore and offshore global climate change. Geological Magazine 151:349–364. classes and J. J. Czaplewski, T. Torsvik, and Finnegan, S., and M. L. Droser. 2005. Relative and absolute abun- dance of trilobites and rhynchonelliform brachiopods across the G. Shephard for help with GPlates. F.F. is Lower/Middle Ordovician boundary, eastern Basin and Range. grateful to C. Syrowatka, J. Starrfelt, Paleobiology 31:480–502.

Harper, D. A. T. 2006. The Ordovician biodiversification: setting an ——. 2012. The Ordovician radiation: macroevolutionary cross- agenda for marine life. Palaeogeography, Palaeoclimatology, roads of the Phanerozoic. Pp. 381–394 in Earth and Life. Springer Palaeoecology 232:148–166. Netherlands, Dordrecht. Harper, D. A. T., C. M. Ø. Rasmussen, M. Liljeroth, R. B. Blodgett, Miller, A. I., and S. R. Connolly. 2001. Substrate affinities of higher Y. Candela, J. Jin, I. G. Percival, J. Rong, E. Villas, and R. Zhan. taxa and the Ordovician radiation. Paleobiology 27:768–778. 2013. Biodiversity, biogeography and phylogeography of Ordo- Miller, A. I., and S. Mao. 1995. Association of orogenic activity with vician rhynchonelliform brachiopods. 1. In D. A. T. Harper and the Ordovician radiation of marine life. Geology 23:305–308. T. Servais, eds. Early Palaeozoic biogeography and palaeogeog- ——. 1998. Scales of diversification and the Ordovician radiation. raphy. Geological Society of London Memoir 38:127–144. Pp. 288–310 in M. L. McKinney and J. A. Drake, eds. Biodiversity Harper, D. A. T., and T. Servais, eds. 2013. Early Palaeozoic bio- dynamics: turnover of populations, taxa, and communities. Col- geography and palaeogeography: towards a modern synthesis. umbia University Press, New York. Geological Society of London Memoir 38:1–4. Müller, R. D., M. Sdrolias, C. Gaina, and W. R. Roest. 2008. Age, Hints, O., A. Delabroye, J. Nõlvak, T. Servais, A. Uutela, and spreading rates, and spreading asymmetry of the world’s ocean Å. Wallin. 2010. Biodiversity patterns of Ordovician marine crust. Geochemistry, Geophysics, Geosystems 9:Q04006. microphytoplankton from Baltica: comparison with other fossil Nichols, J. D., and K. H. Pollock. 1983. Estimating taxonomic diver- groups and sea-level changes. Palaeogeography, Palaeoclimat- sity, extinction rates, and speciation rates from fossil data using ology, Palaeoecology 294:161–173. capture–recapture models. Paleobiology 9:150–163. Holland, S. M., and M. Christie. 2013. Changes in area of shallow Novack-Gottshall, P. M., and A. I. Miller. 2003. Comparative geo- siliciclastic marine habitat in response to sediment deposition: graphic and environmental diversity dynamics of gastropods and implications for onshore–offshore paleobiologic patterns. Paleo- bivalves during the Ordovician radiation. Paleobiology 29:576–604. biology 39:511–524. Paris, F., A. Achab, E. Asselin, C. Xiao-Hong, Y. Grahn, J. Nõlvak, Holland, S. M., and M. E. Patzkowsky. 1996. Sequence stratigraphy O. Obut, J. Samuelsson, N. Sennikov, M. Vecoli, J. Verniers, and long-term paleoceanographic change in the Middle and W. Xiao-feng, and T. Winchester-Seeto. 2004. Chitinozoans. Upper Ordovician of the eastern United States. In B. J. Witzke, Pp. 294–311 in Webby et al. 2004. G. A. Ludvigson, and J. Day, eds. Paleozoic sequence stratig- Pope, M. C., and J. B. Steffen. 2003. Widespread, prolonged late raphy: views from the North American craton. Geological Society Middle to Late Ordovician upwelling in North America: a of America Special Paper 306:117–129. Geological Society of proxy record of glaciation? Geology 31:63–66. America, Boulder, Colo. Pradel, R. 1996. Utilization of capture–mark–recapture for the study Jablonski, D. 2005. Evolutionary innovations in the fossil record: of recruitment and population growth rate. Biometrics 52:703–709. the intersection of ecology, development, and macroevolution. Rasmussen, C. M. Ø., J. Hansen, and D. A. T. Harper. 2007. Baltica: a Journal of Experimental Zoology 304B:504–517. mid Ordovician diversity hotspot. Historical Biology 19:255–261. Jablonski, D., J. J. Sepkoski, D. J. Bottjer, and P. M. Sheehan. 1983. Rasmussen, C. M. Ø., C. V. Ullmann, K. G. Jakobsen, A. Lindskog, Onshore–offshore patterns in the evolution of Phanerozoic shelf J. Hansen, T. Hansen, M. E. Eriksson, A. Dronov, R. Frei, C. Korte, communities. Science 222:1123–1125. A. T. Nielsen, and D. A. T. Harper. 2016. Onset of main Phanero- Kröger, B. 2017. Changes in the latitudinal diversity gradient during zoic marine radiation sparked by emerging Mid Ordovician the great Ordovician biodiversification event. Geology 46:44–47. icehouse. Scientific Reports 6:18884. Kröger, B., and K. Lintulaakso. 2017. RNames, a stratigraphical Rosenzweig, M. L. 1995. Species diversity in space and time. Cam- database designed for the statistical analysis of fossil occur- bridge University Press, Cambridge. rences—the Ordovician diversification as a case study. Palaeonto- Roy, K., D. Jablonski, J. W. Valentine, and G. Rosenberg. 1998. Mar- logia Electronica 20.1.1T:1–12. ine latitudinal diversity gradients: tests of causal hypotheses. Laake, J. L. 2013. RMark: An R Interface for Analysis of Capture- Proceedings of the National Academy of Sciences USA Recapture Data with MARK. AFSC Processed Report 2013-01. 95:3699–3702. Alaska Fisheries Science Center, NOAA, National Marine Fisher- Saltzman, M. R., and S. A. Young. 2005. Long-lived glaciation in the ies Service, Seattle, Wash. Late Ordovician? Isotopic and sequence-stratigraphic evidence Lindskog, A., M. M. Costa, C. M. Ø. Rasmussen, J. N. Connelly, and from western Laurentia. Geology 33:109–112. M. E. Eriksson. 2017. Refined Ordovician timescale reveals no Schmitz, B., M. Tassinari, and B. Peucker-Ehrenbrink. 2001. A rain link between asteroid breakup and biodiversification. Nature of ordinary chondritic meteorites in the early Ordovician. Earth Communications 8:14066. and Planetary Science Letters 194:1–15. Liow, L. H., and J. D. Nichols. 2010. Estimating rates and probabil- Schmitz, B., D. A. T. Harper, B. Peucker-Ehrenbrink, S. Stouge, ities of origination and extinction using taxonomic occurrence C. Alwmark, A. Cronholm, S. M. Bergström, M. Tassinari, and data: capture-mark-recapture (CMR) approaches. In J. Alroy W. Xiaofeng. 2008. Asteroid breakup linked to the great Ordovi- and G. Hunt, eds. Quantitative methods in paleobiology. cian biodiversification event. Nature Geoscience 1:49–53. Paleontological Society Papers 16:81–94. Schwarz, C. J., and A. N. Arnason. 1996. A general methodology Liow, L. H., T. Reitan, and P. G. Harnik. 2015. Ecological interac- for the analysis of capture–recapture experiments in open popu- tions on macroevolutionary time scales: clams and brachiopods lations. Biometrics 52:860–873. are more than ships that pass in the night. Ecology Letters Sepkoski, J. J. 1991. A model of onshore–offshore change in faunal 18:1030–1039. diversity. Paleobiology 17:58–77. MacArthur, R. H., and E. O. Wilson. 1967. The theory of island ——. 1997. Biodiversity: past, present, and future. Journal of biogeography. Princeton University Press, Princeton, N.J. Paleontology 71:533–539. Miller, A. I. 1997a. Comparative diversification dynamics among Sepkoski, J. J., R. K. Bambach, D. M. Raup, and J. W. Valentine. palaeocontinents during the Ordovician radiation. Geobios 1981. Phanerozoic marine diversity and the fossil record. Nature 20:397–406. 293:435–437. ——. 1997b. Dissecting global diversity patterns: examples from Servais, T., and D. A. T. Harper. 2018. The great Ordovician biodi- the Ordovician radiation. Annual Review of Ecology and versification event (GOBE): definition, concept and duration. Systematics, 28:85–104. Lethaia:1–14. ——. 2004. The Ordovician radiation: toward a new global synthe- Servais, T., D. A. T. Harper, J. Li, A. Munnecke, A. W. Owen, and P. sis. Pp. 380–388 in Webby et al. 2004. M. Sheehan. 2009. Understanding the great Ordovician

biodiversification event (GOBE): influences of paleogeography, biodiversification? Evidence from conodont thermometry. paleoclimate, or paleoecology? GSA Today 19:4–10. Science 321:550–554. Sprinkle, J., and T. E. Guensburg. 1995. Origin of echinoderms in Trubovitz, S., and A. L. Stigall. 2016. Synchronous the Paleozoic evolutionary fauna: the role of substrates. Palaios diversification of Laurentian and Baltic rhynchonelliform 10:437–453. brachiopods: implications for regional versus global triggers ——. 2004. Crinozoan, blastozoan, echinozoan, asterozoan and of the great Ordovician biodiversification event. Geology homalozoan echinoderms. Pp. 266–280 in Webby et al. 2004. 44:743–746. Tomašových, A., S. Dominici, M. Zuschin, and D. Merle. 2014. Valentine, J. W. 2009. Overview of marine biodiversity. Pp. 3–28 in Onshore–offshore gradient in metacommunity turnover emerges J. D. Witman and K. Roy, eds. Marine macroecology. University only over macroevolutionary time-scales. Proceedings of the of Chicago Press, Chicago. Royal Society of London B 281:20141533. Webby, B. D., F. Paris, M. L. Droser, and I. G. Percival, eds. 2004. Torsvik, T. H., and L. R. M. Cocks. 2016a. Ordovician. Pp. 101–123 The great Ordovician biodiversification event. Columbia Univer- in Earth history and palaeogeography. Cambridge University sity Press, New York. Press, Cambridge. White, G. C. 2016. MARK: mark and recapture parameter estima- ——. 2016b. Tectonic units of the Earth. Pp. 38–76 in Earth history tion. http://www.phidot.org/software/mark. and palaeogeography. Cambridge University Press. Zhan, R., and D. A. T. Harper. 2006. Biotic diachroneity during the Trotter, J. A., I. S. Williams, C. R. Barnes, C. Lécuyer, and R. Ordovician radiation: evidence from South China. Lethaia S. Nicoll. 2008. Did cooling oceans trigger Ordovician 39:211–226.