3. Macroevolutionary rate metrics

A running curve of standing richness through the entire 74 Ma of the graptoloid clade is readily derived from the scaled and calibrated composite sequence which is, in effect, a global biostratigraphic range chart of 2046 species. The composite levels of FA and LA events (range-end events) are equivalent to sampling levels for taxon range ends in a highly resolved stratigraphic section. Because levels in the CONOP calibrated composite are spaced only 37 ka apart on average and there are, on average, only 1.78 range-end events per level, we have a record of graptoloid species ranges with an exceptionally high temporal resolution. Assuming that the recorded FA and LA events represent species origination and extinctions events, we can regard the record as a literal one in which there is no significant species turnover between levels. Clearly we cannot deny the possibility that a species’ true first appearance did not occur up to 37 ka earlier than we have recorded it, and that no regional biostratigraphic section has detected it; such possibilities will make no significant difference to our recorded originations and extinctions through time, however. We can therefore regard all species recorded at any level as boundary-crossers (sensu Foote, Reference Foote2000). This approach avoids the troublesome biases introduced by time-binning the data (Foote, Reference Foote1994, Reference Foote2000; Foote & Miller, Reference Foote and Miller2007; Alroy, Reference Alroy, Alroy and Hunt2010). Although the species richness curve shown here (Figs 3, 4) is a raw (‘as-measured’) curve, Sadler et al. (Reference Sadler, Cooper and Melchin2011) have shown that the main topographic features of the curve are robust to a variety of tests for sampling completeness. The main features survive rarefaction even at low levels of resampling (50 or 100 samples).

Figure 3. Per capita instantaneous extinction q, origination p and turnover p+ q rate curves (species per Lmy, 1 Ma smoothing) and species richness (unbinned) for the entire graptoloid clade. δ¹³C (carbonate) curves and stage slices are from Cramer et al. (Reference Cramer, Brett, Melchin, Männik, Kleffner, McLaughlan, Loydell, Munnecke, Jeppsson, Corradini, Brunton and Saltzman2011) and Bergström et al. (Reference Bergström, Chen, Gutéirrez-Marco and Dronov2009) and are correlated via graptolite and conodont zones. Positive isotope excursions associated with extinction rate peaks and/or low or falling richness are in darker-shaded bands; those with rising or peak richness are in lighter-shaded bands.

Figure 4. Late Ordovician – Middle Silurian interval spanning the end-Ordovician mass extinction, Llandovery recovery and two of the Silurian major carbon isotope excursions. (a) Unbinned species richness (Sadler et al. Reference Sadler, Cooper and Melchin2011). (b) Per capita instantaneous extinction rate (q, black line) and origination rate (p, grey line) with a sliding 0.5 Ma bin for rate determination. Evolutionary rates are expressed as per Lmy rates. (c) δ¹³C_carb‰ from Ainsaar et al. (Reference Ainsaar, Kaljo, Martma, Meidla, Männik, Nõlvak and Tinn2010) and Cramer et al. (Reference Cramer, Brett, Melchin, Männik, Kleffner, McLaughlan, Loydell, Munnecke, Jeppsson, Corradini, Brunton and Saltzman2011). (d) Inset enlargement with the Sheinwoodian and Homerian isotope excursions (shaded) and the extinction rate in higher resolution (0.25 Ma sliding bin); ±1σ limits shown. Dotted line: δ¹³C curve reconstructed from unaltered brachiopod shell calcite from the Silurian of Gotland (compiled after Samtleben et al. Reference Samtleben, Munnecke, Bickert and Pätzold1996, Reference Samtleben, Munnecke and Bickert2000; Munnecke et al. Reference Munnecke, Samtleben and Bickert2003; Cramer et al. Reference Cramer, Loydell, Samtleben, Munnecke, Kaljo, Männik, Martma, Jeppsson, Kleffner, Barrick, Johnson, Emsbo, Joachimski, Bickert and Saltzman2010). (e) Same as (d) for the late Katian and Hirnantian excursions and extinction rate curve. Dotted line: whole-rock δ¹³C curve from the Hirnantian of Mirni Creek (NE Russia). The peak of the HICE plots well within the persculptus zone (Kaljo & Martma, Reference Kaljo, Martma, Gutiérrez-Marco, Rábano and García-Bellido2011). g – graptolite extinction event; c, c⁻ – carbon isotope event; S – insufficient resolution.

Unlike taxon richness, origination and extinction rates are inherently time-binned concepts. We must determine these rates as within-bin averages with consequent loss of resolution; because the age distribution of evolutionary events within the bin is known in high resolution, there is no need to estimate it. Accordingly, we report the rate data in 1 Ma bins, or shorter-duration bins for higher resolution (0.5 Ma, 0.25 Ma) and we have the freedom to use a sliding bin, centred at each of the 2031successive event levels in the composite. The rate is therefore recalculated at every level producing, in effect, an instantaneous rate curve and a means to scan the data for the most precise placement of turning points in the rate curves. The instantaneous per-capita rates for origination (p), extinction (q) and turnover (p + q) are scaled to lineage length, normally expressed as per lineage million years (Lmy) following Raup (Reference Raup1985) and Foote & Miller (Reference Foote and Miller2007). The number of extinctions N(e) or originations N(o) in any time interval t is divided by the sum of all lineage segments that lie within the interval, in millions of years. Lmy thus relates an extinction (or origination) event to the aggregate time during which all lineages were exposed to extinction (or speciation) within the interval. For the extinction rate in time interval t: q_t = N(e) _t /(Lmy) _t . Similarly, for origination rate in time interval t: p_t = N(o) _t /(Lmy) _t . Species turnover rate is simply q_t+ p_t .

The results are plotted against the δ¹³C_carb data summarized in Bergström et al. (Reference Bergström, Chen, Gutéirrez-Marco and Dronov2009) for the Ordovician and in Cramer et al. (Reference Cramer, Brett, Melchin, Männik, Kleffner, McLaughlan, Loydell, Munnecke, Jeppsson, Corradini, Brunton and Saltzman2011) for the Silurian (Figs 3, 4). Numerous studies have proven δ¹³C_carb to be a reliable chemostratigraphic proxy. In contrast, the carbon isotopic composition of organic matter δ¹³C_org depends on a variety of factors such as light intensity, pH of ambient seawater, organism cell geometry, heterotrophic reworking, metabolic effects and species composition of the phytoplankton community (Hinga et al. Reference Hingaga, Arthur, Pilson and Whitaker1994; Hayes et al. Reference Hayes, Strauss and Kaufman1999; Bickert, Reference Bickert, Schulz and Zabel2006); it therefore often exhibits more ‘noise’ than contemporaneous δ¹³C_carb curves (e.g. Cramer & Saltzman Reference Cramer and Saltzman2007). Furthermore, there is much less published δ¹³C_org data than for δ¹³C_carb.

4. Graptoloid evolutionary rate dynamics

The rates and species richness curves are plotted alongside the generalized global δ¹³C curve in Figure 3. Graptoloid species richness is estimated from our species list of 2094 species, the best-supported graptoloid richness estimate yet published (Sadler et al. Reference Sadler, Cooper and Melchin2011). The impact of an extinction event depends on both the intensity of the event (the per Lmy rate) and the duration of the event; impact therefore depends on the area under the curve. Note that richness falls when the extinction rate q is greater than the origination rate p, whether or not one or both rates are increasing, decreasing or remaining stationary. The converse holds for richness increase.

It can be seen from Figure 3 that graptoloid evolutionary rates through most of the Ordovician were relatively low and of low amplitude compared with the Silurian, during which rates were higher and more volatile. The transition from one state to the other takes place not at the period boundary but in the Katian Stage. In higher resolution (Fig. 4), the extinction rate can be seen to change between c. 446.5 Ma and 451 Ma but the origination and turnover rates change at c. 447 Ma. If we ignore the extreme rates values at the beginning (Tremadocian) and end (Lochkovian) of the clade where richness is very low (fewer than 10 species), and take the boundary between an Ordovician pattern and a Silurian pattern at 447 Ma, we can contrast the evolutionary dynamics of the two periods. For convenience we refer to that part of the Ordovician prior to the late Katian as pK-Ordovician, and to the late Katian to end Silurian as K-Silurian.

Mean species turnover rate for the K-Silurian (1.66/Lmy) is twice that for the pK-Ordovician (0.86/Lmy) and this contrast is reflected in the extinction and origination rates from which the turnover rate is derived. The tempo of graptoloid evolution in the K-Silurian is therefore significantly higher than in the pK-Ordovician. The median duration of a pK-Ordovician species is 1.27 Ma, significantly greater than that of a K-Silurian species (0.69 Ma; Mann Whitney-U p <0.001), reflecting the higher turnover rates. Not only is the mean intensity of evolutionary turnover higher in the K-Silurian, it fluctuates more strongly. During the pK-Ordovician, rate curve excursions were of much lower amplitude (σ = 0.72 for the K-Silurian and 0.33 for the pK-Ordovician; p <0.001; df = 1).

Near the beginning and end of the graptoloid clade, where species richness is very low (10 species or less), the loss (or addition) of just a few species results in extreme values in the proportional rate curves. One of these extremes in the Tremadocian (Tr1) nearly extinguished the clade altogether which, at that time, was represented only by the stem group Anisograptidae; richness dropped to 2 or 3 species in our dataset.

A sustained period of moderately elevated origination rate in the Floian (Fl1, early Bendigonian) produced a rapid increase in species richness of the clade (Floian radiation), which was maintained at 40–100 species throughout the remaining Ordovician. The expansion phase of the Graptoloidea (486–473 Ma) was exponential rather than logarithmic or linear (R ² = 0.8672, R ² = 0.7962 and R ² = 0.7941, respectively) and continued until near-peak richness levels were reached, indicating an adaptive radiation of the clade. The increase was driven by the expansion of the ‘Dichograptidae’ (didymograptids, Tetragrapta) and Sinograpta (includes sigmagraptids). Speculating on the driving mechanisms, Sadler et al. (Reference Sadler, Cooper and Melchin2011) pointed out that the expansion does not seem to have been triggered by breaking through a functional threshold or an evolutionary innovation such as new body plans, thecal types, feeding mechanisms, rhabdosome designs or proximal development types. Rather, there was a proliferation of species using the available designs, for example pendent and extensiform didymograptids, phyllograptid designs and horizontal dichotomous, progressive and monoprogressive branching types. Although some genera such as Expansograptus (which proliferated in this interval) undoubtedly need rationalizing to remove species based on morphotypes rather than populations, it is unlikely that the observed pattern of expansion could be entirely, or even largely, explained by this. Work in progress by the authors indicates that the Ordovician expansion in species richness of graptoloids was facilitated not only by a sustained and moderately elevated origination rate but by an increase in median longevity of species that was maintained through the pK-Ordovician at a level twice that of the Silurian (see Section 7). It was also facilitated by the extinction rate remaining at background levels.

Many marine invertebrate groups radiated during the Early and Middle Ordovician during the Great Ordovician Biodiversification Event (Webby et al. Reference Webby, Droser and Paris2004) and it is likely that, in large part, the graptolites responded to stimuli that were global in extent and pan-taxonomic in their influence. Among these is likely to have been the availability of trophic resources. The marine microphytoplankton, as represented by the acritarchs, diversified through the Late Cambrian and Early Ordovician, peaking in the Darriwilian (Li et al. Reference Le Heron2007; Servais et al. Reference Servais, Lehnert, Li, Mullins, Munnecke, Nützel and Vecoli2008, Reference Servais, Owen, Harper, Kröger and Munnecke2010), suggesting an expanding food resource for the graptoloids (Rigby, Reference Rigby1991; Cooper et al. Reference Cooper, Rigby, Loydell and Bates2012). The microphytoplankton and bacterioplankton proliferate when nutrient levels rise such as by nutrient transfer, a modern analogue for which is found in upwelling zones (Finney & Berry, Reference Finney and Berry1997; Cooper et al. Reference Cooper, Rigby, Loydell and Bates2012). The expansion of the graptoloids is therefore likely to be linked to extrinsic factors such as ocean circulation, biogeochemistry and, ultimately, to global climatic regime.

Prior to the Katian, the main evolutionary event was the Darriwilian Dw3 extinction and richness minimum, during which the Dichograptidae were severely reduced in species numbers and the isograptids and Sinograpta were terminated. The event was preceded in the late Dw1 – early Dw2 by a smaller extinction event which depleted the standing species richness by c. 30%. During the Darriwilian the Diplograptina, represented by the family-level groups Climacograptoidea and Diplograptidae, became established globally and, together with the Dicranograptoidea and Lasiograptidae, dominated graptolite communities during the Late Ordovician. There was therefore a major re-organization of the graptoloid clade during the Darriwilian, strongly assisted by the Darriwilian extinction event. A substantial extinction event at the Katian Ka1–Ka2 boundary was largely offset by elevated origination rate and had relatively little impact on species richness, but coincides with the demise of the Glossograpta. The extirpation of normalograptids from the Laurentian tropics at this time was followed, during the late Katian and Hirnantian, by their re-invasion and dramatic expansion when they replaced the diplograptines (Goldman et al. Reference Goldman, Mitchell, Melchin, Fan, Wu and Sheets2011). The Ka1–Ka2 extinction event may represent a second significant pK-Ordovician event, or was possibly a precursor to the K-Silurian pattern.

With the exception of the Darriwilian and Katian extinction events, the pK-Ordovician pattern suggests that background rates of extinction and turnover prevailed; the K-Silurian pattern of a highly volatile evolutionary dynamic with recurring sharp spikes in the extinction and origination rates however suggests sharp environmental perturbations. Following the Katian–Hirnantian mass extinction (discussed in Section 6), at least six pronounced graptoloid extinction events punctuate the Silurian. Most have been previously recognized (e.g. by Urbanek, Reference Urbanek1993; Štorch, Reference Štorch1995; Melchin et al. Reference Melchin, Koren’, Štorch, Landing and Johnson1998; Loydell, Reference Loydell2007) although our zonal assignment of some of the events differs slightly from that previously reported, mainly because of the different method of zone boundary recognition used here (see discussion in Sadler et al. Reference Sadler, Cooper and Melchin2009).

The Rhuddanian (Rh2–3) extinction event, Cystograptus vesiculosus to Coronograptus cyphus zones, was first recognized by Melchin et al. (Reference Melchin, Koren’, Štorch, Landing and Johnson1998) and referred to as the ‘Late Parakidograptus acuminatus event’. The extinction affected the Normalograptidae, Dimorphograptidae and Neodiplograptidae. The extinction peak was almost entirely offset by a coincident origination rate peak resulting in a high species turnover, noted by Melchin et al. (Reference Melchin, Koren’, Štorch, Landing and Johnson1998, Reference Melchin, Mitchell, Nacyk-Cameron, Fan and Loxton2011), and had little impact on overall species richness. The event triggered a diversification within the Monograptidae and Dimorphograptidae clades (Koren’ & Bjerreskov, Reference Koren’ and Bjerreskov1999).

An Aeronian (Ae2–3) extinction event in the D. argenteus to S. sedgwicki zones corresponds to the L. convolutus and S. sedgwicki event of Melchin et al. (Reference Melchin, Koren’, Štorch, Landing and Johnson1998) and Štorch (Reference Štorch1995) although our data suggest that, in a global context, it may have started somewhat earlier than generally recognized. The event resulted in a severe but brief drop in species richness in the sedgwicki zone. Retiolitids, including the Petalolithinae, and Monograptidae all declined in species richness. ‘Stressed’ marine environmental conditions are inferred by Štorch & Frýda (2012).

In the early Telychian (Te1–2) late S. turriculatus and M. crispus zones, Loydell (Reference Loydell1994) and Štorch (Reference Štorch1995) recognized the Utilis Event. This event, of moderate intensity, is significant in that it marks the beginning of a long decline in global species richness. The spike in origination rate in the guerichi zone accords with previous work, but the high levels of species richness through the turriculatus zone is at variance with some regional studies (Loydell, Reference Loydell1994; Storch, 1994). Biserial graptoloids were nearly eliminated, with a few survivors among the retiolitids, normalograptids and petalolithids (Štorch, Reference Štorch1995).

An extinction event recognized in the Cyrtograptus lapworthi zone of Bohemia (Štorch, Reference Štorch1995), Wales (Loydell, Reference Loydell2007) and Arctic Canada (Melchin, Reference Melchin, Koren’, Štorch, Landing and Johnson1998) shows only as a small spike in the extinction rate in our global data. However, there is a marked depletion in species richness in the late C. lapworthi and C. insectus zones.

The early Sheinwoodian (Sh1) extinction event, Cyrtograptus murchisoni to early M. riccartonensis zones, is a widely recognized event (Koren’, Reference Koren’1987; Kaljo et al. Reference Kaljo, Boucot, Corfield, Herisse, Koren’, Kříž, Männik, Märss, Nestor, Shaver, Siveter, Viira and Walliser1995; Štorch, Reference Štorch1995; Melchin et al. Reference Melchin, Koren’, Štorch, Landing and Johnson1998; Loydell, Reference Loydell2007) of an intensity exceeding that of the Hirnantian Stage (Fig. 2). Global species richness dropped sharply by 50% to 20 species, matching the Hirnantian minimum. Many lineages were terminated (Štorch, Reference Štorch1995) including Barrandeograptus, Mediograptus and the Cyrtograptus murchisoni group. According to the correlations of Loydell et al. (Reference Loydell, Männik and Nestor2003) and Loydell (Reference Loydell2007), the well-known Ireviken conodont extinction event in Gotland (Jeppsson, Reference Jeppsson, Brett and Baird1997) coincides with this graptolite event. Some species such as Monoclimacis vomerina group and Monograptus priodon were exterminated from one or more regions only to reappear several zones later (Lazarus taxa; Štorch, Reference Štorch1995). These local extermination events contribute to some regional extinction counts but are not recorded as extinction events in our composite sequence.

In the early Homerian (Ho1), Cyrtograptus lundgreni zone, the ‘Große Krise’ of Jaeger (Reference Jaeger1991) has been recognized by many others (Koren’, Reference Koren’1991; Lenz, Reference Lenz1993; Štorch, Reference Štorch1995; Gutiérrez-Marco et al. Reference Gutiérrez-Marco, Lenz, Robardet and Picarra1996; Melchin et al. Reference Melchin, Koren’, Štorch, Landing and Johnson1998; Lenz & Kozlowska-Dawidziuk, Reference Lenz2001; Loydell, Reference Loydell2007) as a major extinction event. Species richness was reduced to its lowest level since the Floian. The event was immediately followed by two, smaller peaks in extinction rate during the late Homerian (Ho2–3; Fig. 4d). The plectograptines and species of the Pristiograptus dubius group were among the few to have survived; these and other lineages including Gothograptus, Saetograptus and graptolites with peculiar apertural modifications, the cucullograptids, then speciated rapidly, generating an exponential rise to a short-lived peak in global richness and a final ‘fling’ for the graptoloids.

During the late Ludfordian to early Pridoli an extended and multi-peaked extinction episode almost terminated the only remaining group, the superfamily Monograptoidea. It was accompanied by a sustained elevated origination rate and, as a consequence, a high species turnover rate. Three main extinction events of successively increasing intensity correspond, in upward sequence, to the Saetograptus leintwardensis, Neocucullograptus kozlowskii and Monograptus transgrediens zones. All three were recognized in the sections of Bohemia and eastern Europe (Urbanek, Reference Urbanek1993; Štorch, Reference Štorch1995). The first event terminated the plectograptines, the second terminated several genera including Neocucullograptus, Bohemograptus and Polonograptus (Manda et al. Reference Manda, Štorch, Slavík, Frýda, Křiž and Tásaryová2012) and the third event removed Pseudomonoclimacis, Pristiograptus and all other species of Monograptus s.l. (Koren’, 1993; Urbanek, Reference Urbanek1993). Global species richness dropped to just 4 species and did not rise above 10 species for the following 9 Ma, after which the Graptoloidea disappeared completely.

It is emphasized that these extinction events and the species richness pattern are global in scope and may differ from the extinction and richness history of local regions. At higher resolution (0.5 Ma and 0.25 Ma sliding windows, Figs 2, 4) several of the extinction rate spikes, including that of the late Katian – Hirnantian, are seen to comprise multiple peaks which are significantly different at the 0.05 level. Throughout the history of the graptoloid clade, all major depletions in richness except for one were caused by elevated extinction rates (Fig. 3). The exception occurs during the Sandbian Sa2 where a pronounced decline in richness was caused by a falling origination rate rather than by elevated extinction.

5. Stable carbon isotope (δ¹³C) anomalies

A contrast between an Ordovician and a Silurian pattern similar to that in graptoloid evolution is seen in the global generalized δ¹³C_carb curve (Fig. 3; Bergström et al. Reference Bergström, Chen, Gutéirrez-Marco and Dronov2009; Cramer et al. Reference Cramer, Brett, Melchin, Männik, Kleffner, McLaughlan, Loydell, Munnecke, Jeppsson, Corradini, Brunton and Saltzman2011; Saltzman & Thomas, Reference Saltzman, Thomas, Gradstein, Ogg, Schmitz and Ogg2012). For more detail we show the curve of Ainsaar et al. (Reference Ainsaar, Kaljo, Martma, Meidla, Männik, Nõlvak and Tinn2010) based on the Baltic sequence, the Kaljo & Martma (Reference Kaljo, Martma, Gutiérrez-Marco, Rábano and García-Bellido2011) curve of the Mirny Creek sequence, Russia and a composite curve based on the Gotland Silurian succession (Fig. 4). The Silurian has been described as a time of extreme instability in the global carbon cycle (Munnecke et al. Reference Munnecke, Samtleben and Bickert2003, Reference Munnecke, Zhang, Liu and Cheng2010; Cramer et al. Reference Cramer, Loydell, Samtleben, Munnecke, Kaljo, Männik, Martma, Jeppsson, Kleffner, Barrick, Johnson, Emsbo, Joachimski, Bickert and Saltzman2010, Reference Cramer, Brett, Melchin, Männik, Kleffner, McLaughlan, Loydell, Munnecke, Jeppsson, Corradini, Brunton and Saltzman2011) and contrasts with the much more subdued excursions through most of the Ordovician. Further, the transition from an Ordovician to a Silurian pattern takes place during the Katian (Saltzman & Young, Reference Saltzman and Young2005), as with the graptolites. Positive isotopic excursions in the Hirnantian, late Aeronian (Ae3), early Telychian (Te2), Sheinwoodian (Sh1–2), Homerian (Ho2–3) and Ludfordian (Lu1–2) all lie at times of depressed or falling graptoloid species richness, most also at times of elevated extinction rate (Figs 3, 4; Loydell, Reference Loydell2007). The Sheinwoodian event has been studied intensively (Munnecke et al. Reference Munnecke, Samtleben and Bickert2003; Cramer et al. Reference Cramer, Loydell, Samtleben, Munnecke, Kaljo, Männik, Martma, Jeppsson, Kleffner, Barrick, Johnson, Emsbo, Joachimski, Bickert and Saltzman2010); conodonts, trilobites, brachiopods, corals and ostracods were exterminated in successive waves of extinction in the Baltic carbonate platform (Ireviken Event) and globally during the latest Telychian (Te5), immediately preceding the onset of both the isotope excursion and the graptolite extinction event (Munnecke et al. Reference Munnecke, Calner, Harper and Servais2010) or during the graptolite extinction event itself (Loydell et al. Reference Loydell, Männik and Nestor2003; Loydell, Reference Loydell2007).

In high resolution (Fig. 4), the extinction of graptolite species can be seen to coincide precisely with the positive isotope excursion during the early Sheinwoodian. There is a similar correlation during the Hirnantian, but here the duration of the extinction event is somewhat longer than that of the isotope event. There is an ongoing debate on the precise correlation of the Hirnantian δ¹³C_carb excursion (HICE) and graptolite biostratigraphy because sections with good biostratigraphic control such as Dob's Linn (Scotland) and the Wangjiawan North section (China) do not contain enough carbonate for δ¹³C_carb measurements (Underwood et al. Reference Underwood, Crowley, Marshall and Brenchley1997; Chen et al. Reference Chen, Rong, Fan, Zhan, Mitchell, Harper, Melchin, Peng, Finney and Wang2006), whereas carbonate-rich successions such as on Anticosti or in S. Sweden are usually very sparsely populated with diagnostic graptolites (Melchin & Holmden, Reference Melchin and Holmden2006; Melchin, Reference Melchin2008; Ainsaar et al. Reference Ainsaar, Kaljo, Martma, Meidla, Männik, Nõlvak and Tinn2010; Bergström et al. Reference Bergström, Lehnert, Calner and Joachimski2012). There are a few exceptions however, such as the Vinini Creek section in central Nevada (Finney et al. Reference Finney, Berry, Cooper, Ripperdan, Sweet, Jacobson, Soufiane, Achab and Noble1999), the Mirny Creek section in SE Russia (Kaljo & Martma, Reference Kaljo, Martma, Gutiérrez-Marco, Rábano and García-Bellido2011) and the Honghuayuan and Daijiagou sections in S. China (Munnecke et al. Reference Munnecke, Zhang, Liu and Cheng2011). The main question is whether the maximum δ¹³C values are correlated with the extraordinarius or with the persculptus zone. Both the Vinini Creek and the Mirny sections show the maximum values clearly positioned within the persculptus zone. At Dob's Linn, the stratotype for the Ordovician–Silurian boundary, the Hirnantian δ¹³C peak measured from organic material, is recorded from strata containing both Metabolograptus extraordinarius and Metabolograptus persculptus and thus belongs to the lower M. persculptus zone (Kaljo et al. Reference Kaljo, Hints, Männik and Nõlvak2008; Delabroye & Vecoli, Reference Delabroye and Vecoli2010). In S. China, the maximum values are reported from strata assigned to the M. extraordinarius zone (Munnecke et al. Reference Munnecke, Zhang, Liu and Cheng2011). However, as M. persculptus has not been found in these sections, a position of the δ¹³C peak within the lower M. persculptus zone cannot be excluded.

During the Homerian (Ho2–3), two maxima in the isotope curve align with two graptolite extinction events, but a larger extinction event in the preceding Ho1 (see review in Cramer et al. Reference Cramer, Condon, Söderlund, Marshall, Worton, Thomas, Calner, Ray, Perrier, Boomer, Patchett and Jeppsson2012) has no matching isotopic event (Fig. 4d). The big isotopic excursion during the Ludfordian (Samtleben et al. Reference Samtleben, Munnecke, Bickert and Pätzold1996, Reference Samtleben, Munnecke and Bickert2000; Jeppsson & Aldridge, Reference Jeppsson and Aldridge2000; Jeppsson et al. Reference Jeppsson, Talent, Mawson, Simpson, Andrew, Calner, Whitford, Trotter, Sandström and Calcon2007; Kaljo et al. Reference Kaljo, Grytsenko, Martma and Motus2007; Kozlowski & Munnecke, Reference Kozlowski and Munnecke2010; Fig. 3) aligns with sharply falling richness. Many of the strong K-Silurian carbon isotope excursions are also associated with extinction events in the shallow marine carbonate facies and with positive shifts in the δ¹⁸O isotope ratio; however the causes of these correlations are uncertain (see Munnecke et al. (Reference Munnecke, Samtleben and Bickert2003, Reference Munnecke, Zhang, Liu and Cheng2010) for a summary).

There are exceptions to the general match of positive isotope excursions with richness depletions and extinction rate peaks. The broad positive isotope deflections in the middle Darriwilian (MDICE) and Aeronian (Ae1) overlap times of high species richness. In the Tremadocian (Tr2, R. manitouensis zone) a small but sharp positive carbon isotopic event aligns with increasing graptoloid species richness (Buggisch et al. Reference Buggisch, Keller and Lehnert2003). With the possible exception of the Darriwilian Dw3 (Turner et al. Reference Turner, Armstrong, Wilson and Makhlouf2012), the much lower magnitude of the pK-Ordovician positive isotope excursions and the fact that three of them (marked by lightly shaded bands in Fig. 3 and by c⁻ in Fig. 4c) lack a corresponding diversity depletion or extinction rate peak suggest that their generating mechanism may be quite different from those of the K-Silurian. Major extinction peaks at the Sheinwoodian–Homerian boundary and the Pridoli (Pr1–Pr2 boundary) have no matching isotope excursion.

Statistical testing of the cross-correlations shows a complex pattern, consistent whether using 0.25 Ma or 0.5 Ma time bins. The 0.25 Ma bin results are quoted here. In the K-Silurian, detrended richness correlates negatively with δ¹³C_carb (Pearson's R = –0.235, p = 0.021 with a 0.5 Ma lag), consistent with the pattern described above. In the pK-Ordovician however, after the Floian expansion, richness correlates positively with δ¹³C_carb (R = 0.340, p < 0.001 with a 0.25 Ma lag). Further, although many of the extinction rate peaks coincide closely with positive isotopic excursions, the overall extinction rate itself does not correlate significantly with carbon isotope ratio in either the Ordovician or Silurian. The same applies for the origination rate.

The general coincidence of major positive isotopic excursions with graptoloid depletions in richness in the K-Silurian is therefore supported and, as both the carbon isotopic composition and graptoloid ecology are intimately linked to oceanic geochemistry, circulation, bio-productivity, sea-level-induced carbonate weathering cycles and surface water temperature, this coincidence suggests a common environmental cause.

6. The Katian–Hirnantian depletion in richness

The spectacular decline in graptoloid species richness during the late Katian and Hirnantian is seen to be a monotonic event and can be dated precisely at 448–446 Ma (Fig. 4). Our data suggest that the crisis was a short-lived event. If we take the widest view (from the beginning of the decline to the return to maximum richness in the late Aeronian), it lasted 8 Ma. If we take the time from average Katian richness levels before the event to average Rhuddanian–Aeronian levels after the event, it was only 3 Ma (Fig. 4). In high resolution (Fig. 4e), we can see that the extinction event alone was only 1.3 Ma. This may be why the Hirnantian extinction does not show prominently in some analyses using a coarser timescale (e.g. Alroy, Reference Alroy2008).

The drop in richness was driven initially during the late Katian by rising extinction rate accompanied by a drop in the origination rate and subsequently in the Hirnantian by the high rate of extinction alone (Fig. 4b). However, whereas species richness declined in a single step, the high resolution extinction rate curves (Fig. 2b, c) show a distinct early peak in the early Ka4 and higher multiple peaks in the Hirnantian. The per million year rate of extinction was higher in the Katian than in the Hirnantian (Fig. 5), and many more graptoloid species were extinguished in the Katian than in the Hirnantian (Fig. 4a).

Figure 5. The number of species going extinct each million years (y-axis) peaked in the late Katian and again in the Rhuddanian. X-axis is age in Ma.

In the later part of the Ordovician, the Graptoloidea comprised two major high-level clades, the Diplograptina and the Neograptina (Melchin et al. Reference Melchin, Mitchell, Nacyk-Cameron, Fan and Loxton2011). In the late Katian (P. pacificus and pre-pacificus zones) species richness reached a peak of 86 at 447.8 Ma, with the expansion of the Diplograptina. As a result of the subsequent depletion in richness, standing species richness dropped to 20 in the Hirnantian (445.82 Ma), a loss of 77% of species in 2 Ma (see also Melchin & Mitchell, Reference Melchin, Mitchell, Barnes and Williams1991; Chen et al. Reference Chen, Melchin, Sheets, Mitchell and Fan2005). The entire Diplograptina were extirpated, including the family-level groups Diplograptidae, Climacograptoidea, Dicranograptoidea as well as the Lasiograptidae, resulting in the near extinction of the clade and a major resetting of the diversity pattern and evolutionary history of graptoloids (Melchin & Mitchell, Reference Melchin, Mitchell, Barnes and Williams1991; Chen et al. Reference Chen, Melchin, Sheets, Mitchell and Fan2005; Finney et al. Reference Finney, Berry and Cooper2007; Bapst et al. Reference Bapst, Melchin, Sheets and Mitchell2012; Melchin et al. Reference Melchin, Mitchell, Nacyk-Cameron, Fan and Loxton2011; Sadler et al. Reference Sadler, Cooper and Melchin2011; Štorch et al. Reference Štorch, Mitchell, Finney and Melchin2011). In high resolution (Fig. 4e), the global graptoloid extinction rate was significantly elevated for c. 1.3 Ma and shows at least two main peaks in the Hirnantian. Two main episodes were found by Chen et al. (Reference Chen, Melchin, Sheets, Mitchell and Fan2005) in the Yangtze Platform sections where the event was estimated to have lasted 0.6–0.9 Ma. A two-strike extinction episode during the Hirnantian is also widely recognized in some benthic fossil groups such as brachiopods (Brenchley et al. Reference Brenchley, Carden and Marshall1995; Rasmussen & Harper, Reference Rasmussen and Harper2011).

7. The Llandovery recovery

Minimum richness was very short lived. The surviving high-level clade, the Neograptina (represented by the Normalograptidae and the Retiolitoidea) which for the previous 20 Ma had remained at relatively low richness, suddenly began to expand, creating a spike in the origination rate during the late Hirnantian. The diversification of the normalograptids followed their re-invasion of low palaeo-latitude regions (Goldman et al. Reference Goldman, Mitchell, Melchin, Fan, Wu and Sheets2011). Species richness immediately began to rise during Aeronian time, recovering to Ordovician peak levels within 5 Ma. The Rhuddanian–Aeronian recovery, although initiated by a peak in the origination rate, was however maintained by a complex interplay of origination and extinction rates. Both rates rose and fell sharply in lockstep in several cycles, with the balance in favour of origination (Fig. 4b). As a result of this strong correlation between the two rate curves, species richness continued to progressively rise through the Rhuddanian and Aeronian despite several sharp extinction events, particularly the Rh2–3 event. Our data indicate a rapid response time; the lag between an extinction rate rise and the ensuing origination increase during the recovery phase (through the Rhuddanian and Aeronian) is less than 0.1 Ma and is c. 0.5 Ma in the Hirnantian (Fig. 4a, b).

The recovery was driven by a dramatic expansion of the Neograptina. Melchin et al. (Reference Melchin, Sadler, Cramer, Gradstein, Ogg, Schmitz and Ogg2012) recognized three phases in the radiation of the Neograptina: an early radiation at the base of the Hirnantian; a second phase starting in the post-glacial latest Hirnantian; and a third phase during the middle Rhuddanian. The neograptine family-level clades Normalograptidae and Neodiplograptidae expanded rapidly, followed by the Retiolitidae and Monograptoidea which began their primary diversification in the third phase.

It is instructive to compare the rate dynamics behind the Llandovery recovery with those of the Floian radiation (for purposes of comparison here taken as 446–441 Ma and 478–473 Ma, respectively). Although the Llandovery recovery records a smaller increase in standing species richness (44 species versus 56 species in the Floian; Table 2), it involved the generation of many more new species; 229 species appeared during the Llandovery recovery with a mean origination rate of 1.043, and 149 species appeared during the radiation of the graptoloid clade during the Floian with a mean origination rate of 0.623. It is the much lower extinction rate during the Floian (0.349 versus 0.863 in the Llandovery) that explains this apparent contradiction, resulting in longer durations of Floian species. Median species duration in the Floian radiation is 1.806 Ma whereas median duration in the Llandovery recovery is 0.863 Ma.

The ecological factors involved in the end-Ordovician extinction and Llandovery recovery of the graptoloid clade have been discussed by Melchin & Mitchell (Reference Melchin, Mitchell, Barnes and Williams1991), Chen et al. (Reference Chen, Melchin, Sheets, Mitchell and Fan2005) and Finney et al. (Reference Finney, Berry and Cooper2007). Bapst et al. (Reference Bapst, Melchin, Sheets and Mitchell2012) found that the rebound in morphologic disparity lagged the rebound in species richness during the early recovery phase in the Hirnantian, and they concluded that unknown and non-ecological factors were controlling this phase of the recovery.

8. Graptoloid extinctions and global climate

In the past two decades the picture of the Ordovician and Silurian climatic development has changed significantly. Oxygen isotope data indicate a long phase of cooling during the Ordovician, with a climax in the Hirnantian (Trotter et al. Reference Trotter, Williams, Barnes, Lecuyer and Nicoll2008). There is however still no agreement on the tropical sea-surface temperatures (Pucéat et al. Reference Pucéat, Joachimski, Bouilloux, Monna, Bonin, Motreuil, Morinière and Hénard2010; see review in Munnecke et al. Reference Munnecke, Calner, Harper and Servais2010) with a wide range of interpretations, for example, from >10 °C warmer than today's oceans (Came et al. Reference Came, Eiler, Veizer, Azmy, Brand and Weidman2007) to c. 10 °C cooler than today for the early Silurian (Giles, Reference Giles2012). Similarly, there is no general agreement on Silurian eustacy (Munnecke et al. Reference Munnecke, Calner, Harper and Servais2010). The amplitudes of the δ¹⁸O and δ¹³C excursions are extremely high compared to the Mezozoic and Cenozoic, hampering the application of Cenozoic models. Nevertheless, most authors agree that the pronounced stable carbon and oxygen isotope fluctuations are not a diagenetic artefact but reflect strong environmental changes. The carbonate carbon isotope record is linked directly to the biosphere and to the global carbon cycle (Saltzman & Thomas, Reference Saltzman, Thomas, Gradstein, Ogg, Schmitz and Ogg2012). According to Saltzman (Reference Saltzman2005), the pre-Katian time of comparatively stable δ¹³C values (Fig. 3) reflects negative feedbacks on bioproductivity in a warm ocean with a low nitrogren:phosphorus ratio in which anoxia led to increased denitrification. During the Sandbian, this system changed to a cooler well-oxygenated phosphorus-limited ocean, which lasted until the Early Devonian.

Graptoloids are inferred to have been a major component of the Early Palaeozoic macrozooplankton (Bulman, Reference Bulman1964; Rickards, Reference Rickards1975; Cooper et al. Reference Cooper, Rigby, Loydell and Bates2012). The habitat of the planktic graptoloids was intimately linked with ocean circulation, density stratification, upwelling systems, nutrient flux and temperature gradients, most probably through their inferred dependency on the marine microphytoplankton and bacterioplankton as a food resource (Rigby, Reference Rigby1991; Cooper et al. Reference Cooper, Rigby, Loydell and Bates2012). These oceanic parameters are in turn closely linked with global climatic regime. It is therefore not surprising that the broad graptolite macroevolutionary pattern matches the broad global climatic pattern, particularly the contrast between the pK-Ordovician and K-Silurian intervals. Our graptoloid evolutionary rates provide strong supporting evidence that this transition took place during the Katian (Saltzman & Young, Reference Saltzman and Young2005; Munnecke et al. Reference Munnecke, Calner, Harper and Servais2010; Finnegan et al. Reference Finnegan, Bergmann, Eiler, Jones, Fike, Eisenman, Hughes, Tripati and Woodward2011).

The late Katian–Hirnantian graptoloid extinction, diversity depletion and mass faunal turnover is clearly part of the global Hirnantian Mass Extinction associated with climatic deterioration and polar ice-sheet growth (Ghienne, Reference Ghienne2003; Le Heron, Reference Le Heron2007), and recognized as one of the ‘big five’ mass extinctions during the Phanerozoic (Sepkoski, Reference Sepkoski, Cooper, Droser and Finney1995; Bambach et al. Reference Bambach, Knoll and Wang2004). The six sharp graptoloid extinction events during the Silurian are of comparable intensity to that of the late Katian – Hirnantian. They suggest that the extreme environmental conditions that brought about the Katian–Hirnantian event recurred through the Silurian. Continental glacial events have been recognized at several levels at least during the early Silurian (Díaz-Martínez & Grahn, Reference Díaz-MartÍnez and Grahn2007; Loydell, Reference Loydell2007), and the oxygen isotopic evidence suggests strong fluctuations in sea-surface temperatures during the K-Silurian (summarized in Munnecke et al. Reference Munnecke, Calner, Harper and Servais2010; Finnegan et al. Reference Finnegan, Bergmann, Eiler, Jones, Fike, Eisenman, Hughes, Tripati and Woodward2011). The graptolite evolutionary rates suggest a means of fine-tuning these events.

However, the precise correlation of an extinction event or diversity decline with its accompanying carbon isotope excursion is not straightforward. Not all extinction events have matching isotope excursions; of those that do, not all align consistently. It has been suggested that isotopic, biotic and climatic events and sea-level change follow a cascading cause-and-effect cycle rather than a simultaneous happening (Samtleben et al. Reference Samtleben, Munnecke and Bickert2000; Cramer & Munnecke, Reference Cramer and Munnecke2008; Cramer et al. Reference Cramer, Loydell, Samtleben, Munnecke, Kaljo, Männik, Martma, Jeppsson, Kleffner, Barrick, Johnson, Emsbo, Joachimski, Bickert and Saltzman2010; Munnecke et al. Reference Munnecke, Delabroye, Servais, Vandenbroucke and Vecoli2012). Further, it is likely that biotic interactions, such as reduced competition for food resources and niche space following environmentally induced population crashes, also influence diversity changes, especially in the K-Silurian.

The cross-correlations discussed suggest that environmental parameters, as represented by δ¹³C_carb, are linked to diversity, as represented by graptoloid species richness, but that this operates positively in the pK-Ordovician and negatively in the K-Silurian. The link between extinction rate and environmental change is less easy to interpret. Many, but not all, of the major extinction rate peaks in the K-Silurian align with isotopic events, but the overall extinction rate does not correlate significantly with δ¹³C. The question needs further investigation. Regardless, graptoloid evolutionary dynamics appear to exhibit a similar sensitivity to climatically forced global changes in ocean circulation and chemistry as planktonic foraminifera in the Mesozoic and Cenozoic (Peters et al. Reference Peters, Kelly and Fraass2013). We predict that climatically forced, sharp environmental perturbations such as glacio-eustatic events (Calner, Reference Calner and Elewa2008) will be found at, or close to, the levels of those graptoloid extinction events associated with diversity declines and positive carbon isotope excursions.