Chapter 24 Late Ordovician zooplankton maps and the climate of the Early Palaeozoic Icehouse

Abstract Chitinozoans and graptolites are the main components of preserved Ordovician zooplankton. As with much of the modern plankton, the ‘first-order’ species distributions of Ordovician plankton reflect water masses defined on the basis of sea surface temperatures. For ‘time slices’ of less than a few million years, zooplankton distribution patterns can be used to infer latitudinal sea surface temperature gradients, key palaeoceanographical boundaries and the position of Ordovician climate belts. Here, using two Late Ordovician time intervals – the early Sandbian and Hirnantian – we review how zooplankton distribution patterns identify Late Ordovician cooling and reflect the development of severe icehouse conditions. Supplementary material: Additional information on methods and material is available at: http://www.geolsoc.org.uk/SUP18670

*Corresponding author (e-mail: thijs.vandenbroucke@univ-lille1.fr) Abstract: Chitinozoans and graptolites are the main components of preserved Ordovician zooplankton. As with much of the modern plankton, the 'first-order' species distributions of Ordovician plankton reflect water masses defined on the basis of sea surface temperatures. For 'time slices' of less than a few million years, zooplankton distribution patterns can be used to infer latitudinal sea surface temperature gradients, key palaeoceanographical boundaries and the position of Ordovician climate belts. Here, using two Late Ordovician time intervals -the early Sandbian and Hirnantian -we review how zooplankton distribution patterns identify Late Ordovician cooling and reflect the development of severe icehouse conditions.
Supplementary material: Additional information on methods and material is available at: http://www.geolsoc.org.uk/SUP18670 The spatial distribution of modern-day marine plankton is controlled by seawater temperature, salinity, food supply and light penetration. These factors, in turn, reflect the prevailing geography and climate, which control ocean circulation patterns and temperature distribution at the surface of the oceans. Understanding past sea surface temperature (SST) distribution is one of the most useful targets for reconstruction in palaeoceanography (Haywood et al. 2005). SSTs help decipher the past behaviour of ocean circulation and ocean heat transport and provide critical input and validation data for numerical climate reconstructions using General Circulation Models (GCMs) (Wefer et al. 2000;Haywood et al. 2005).
A 'best practice' example of how fossil plankton can be used to help reconstruct ancient climates is the 'Pliocene Research, Interpretation and Synoptic Mapping' (PRISM) palaeoenvironmental reconstruction program of the US Geological Survey. It consists of several generations and series of global-scale datasets of marine fossil distributions (foraminiferans, diatoms, etc.) and deduced climate proxy data (SST using transfer functions, modern analogue techniques, but also Mg/Ca ratios, etc.) that are in turn used to improve GCM reconstruction (Dowsett 2007;Dowsett et al. 2010). PRISM also includes extensive continental vegetation datasets that have enabled model -data comparisons (e.g. Salzmann et al. 2008).
The use of fossil plankton to reconstruct late Cenozoic palaeoclimates has the advantage of using species that are either still extant or are closely related to modern ones, even if ecological tolerances of these taxa most likely change with time. The challenge for Palaeozoic workers now is to try to reconstruct these ecological tolerances for their extinct fossil groups and species where there are no close descendants.

Graptolites
Graptolites were an important part of the Ordovician to Early Devonian macro-zooplankton and various models using palaeodepth, palaeolatitude and water mass-specific controls to explain species distribution have been developed (e.g. Cooper et al. 1991;Finney et al. 2007). Graptolites, in essence, resolve into three broad ecological groups based on their facies associations (see Cooper et al. 1991 andCooper &Sadler 2010 for an overview and more details): (a) a group found in distal, deeper water, outer shelf to continental slope facies; (b) a low-diversity group confined to proximal, shallower water, mid to inner shelf deposits; and (c) a group that ranges across these settings, from the inner shelf into outer shelf to slope facies. These groups were distinguished on the basis of the benthic lithofacies in which the graptolites accumulated, and comprise a mix of assemblages within the overlying water column. Two hypotheses exist to explain these groupings: the first emphasizes depth stratification as the primary control on the species groupings ('biotopes'), which were indicative of discrete water masses at various depths (Cooper et al. 1991;Cooper 1998;Cooper & Sadler 2010, and references therein). In this model, the first group represents a deep-water biotope and the third one an epipelagic biotope ( Fig. 24.1). Recent evidence in favour of such a scenario includes Cooper & Sadler's (2010) constrained optimization experiments that showed differences in mean species duration between species of their epipelagic and those of their deep-water groups: species of the deep-water group generally have much shorter biostratigraphical ranges than the ones from the epipelagic groups. According to Cooper & Sadler (2010), this difference in mean species duration between the groups reflects environmental differences, and implies that these groups could not have shared the same habitat (i.e. could not have lived at the same depth). The second hypothesis emphasizes horizontal on-shore/off-shore differentiation as the main control on the graptolite environmental groupings (e.g. Finney 1984Finney , 1986Finney & Berry 1997;Williams et al. 2003;Mitchell et al. 2008), and suggests that groups (a) -(c) inhabited water masses at largely the same depth, but at different areas on a shelf to basin transect ( Fig. 24.1). Finney & Berry (1997), for example, identified 'margin dwellers' (in contrast to the 'deep-water' group) and cratonic invaders (in contrast to the epipelagic group). Cooper et al. (1991) also showed that graptolites had broad 'provincial' distributions, identifying a low-latitude Pacific and a high-latitude Atlantic province for the 'Arenig' (also see Skevington 1974). Vandenbroucke et al. (2009) developed this idea, and in a review of global species distribution for the early Sandbian showed that SST was the primary ecological control on graptolite distributions. Early Sandbian graptolite species could be grouped into planktonic provinces (in their paper called biotopes) that reflected climate belts. Vandenbroucke et al. (2009) noted that, within the ocean mixed layer, apparently deeper and shallowwater species (of groups (a) and (c)) displayed a similar degree of latitudinal differentiation, implying that any depth stratification was within the ocean mixed layer (i.e. above the thermocline) and was controlled by primary ecological factors other than temperature. Difference in depth between these two groups seems to have been small, if present at all.

Chitinozoans
Chitinozoans are very useful in Ordovician to Devonian biostratigraphy and inter-basin correlations; a relatively high-resolution record of their occurrences has been established in many parts of the world (see references in Paris 1996;Paris et al. 1999). However, and in comparison with other fossil groups, they still are much less frequently used for palaeoenvironmental studies and reconstructions, and the ecology of the group is much less well understood than their temporal distribution.
The chitinozoan organic-walled microfossil egg cases and/ or their enigmatic metazoan parent organisms were part of the epipelagic zooplankton. As with the graptolites, Late Ordovician species distributions were parallel to palaeolatitude and SST has been hypothesized as the primary ecological factor controlling their distribution, subsequently potentially modified by palaeocurrents (Paris et al. 1999;Vandenbroucke et al. 2010a, b). The main evidence for this interpretation is summarized here: (1) The distribution of chitinozoan species is largely faciesindependent. This was recently quantified by Vandenbroucke et al. (2010a) for an early Sandbian time slice. Data from this interval show that chitinozoan species occur in shelf to offshelf deposits and that the majority of species occur in both shelf and off-shelf settings. This indicates that the chitinozoan animal lived within the shallow 'mixed layer' of the ocean, independent of any seafloor facies control.
(2) Chitinozoans occur in widespread anoxic deposits, such as the early Silurian black graptolitic 'hot' shales of northern Gondwana and the widespread Ordovician-Silurian anoxic facies in Scotland and Wales. Fossil groups occurring in these deposits must have lived high in the water column, and must have had an epipelagic/nektonic mode of life, considering the lethal anoxic seafloor conditions over large areas.
(3) Chitinozoans occupy a series of latitudinally discrete biotopes, reflecting the latitudinal equator-to-pole temperature gradient and the distribution of climate belts. Chitinozoans must therefore have lived in the topmost 'mixed layer' of the oceans as, below the thermocline, at the base of this layer, there is little latitudinal variation in seawater temperature.
The Ordovician was a period of increasing continental dispersion (Cocks & Torsvik 2002). Historically, chitinozoan research has been focused on North America (Laurentia), NE Europe (the circum-Iapetus palaeocontinents of Baltica and Avalonia) and North Africa and the Middle East (Gondwana). Each had an at times endemic chitinozoan fauna that is reflected in the development of separate biozonal schemes for each palaeocontinent (Paris et al. 1999). As chitinozoans were epipelagic and each of these palaeocontinental areas was located at different palaeolatitudes, it is likely that these chitinozoan 'provinces' largely reflect SST and palaeoclimate belts. In contrast, the Ordovician continental dispersion also reflected increasing endemism of the shelf benthos (see elsewhere in this volume). A number of geographical provinces of benthic organisms living on the shelves of the individual continents provide a good palaeogeographical tool for continent identification. It is important not to confuse the meaning of, and main control on, benthic (palaeocontinent) and planktonic (SST) provinces, as each has its value for better understanding of the Ordovician world.
Here graptolite group found in distal, deeper water, outer shelf to continental slope facies, low diversity graptolite group confined to proximal, shallower water, mid to inner shelf deposits graptolite group that ranges from the inner shelf, into outer shelf to slope facies shelf basin our zooplankton distribution patterns, and evaluate how these influence previous reconstructions of Ordovician plankton distribution.
The 'BugPlates' palaeogeographical reconstruction 'BugPlates' is a modified version of a GIS-orientated palaeogeographical reconstruction software package, which interfaces with fossil databases and which has been developed at the Centre for Geodynamics at the Geological Survey of Norway (Torsvik & Cocks 2009). Reconstructions are based on palaeomagnetic and (predominantly benthonic) palaeobiogeographical data. The base maps used in the reconstructions are updated versions of those published by Cocks & Torsvik (2002. The advantages of using 'BugPlates' are (a) that these are the most recently revised maps available for the Ordovician, (b) that the data they are based on are available as case studies (Cocks & Torsvik 2005, 2007 and (c) that it is in a user-friendly, digital format. Vandenbroucke et al. (2009Vandenbroucke et al. ( , 2010a compiled graptolite and chitinozoan presence/absence data for an early Sandbian time slice, which is equivalent to the entire temporal range of the graptolite Nemagraptus gracilis (see Vandenbroucke et al. 2009, for further discussion), and not to the gracilis graptolite Biozone as variously defined in different palaeogeographical regions (see Williams et al. 2004 for a fuller explanation). The Sandbian data compilation of Vandenbroucke et al. (2009Vandenbroucke et al. ( , 2010a predates the publication of the 'BugPlates' palaeogeography and their data was plotted using the PALEOMAP reconstruction. The Sandbian data are here replotted using 'BugPlates'. Vandenbroucke et al. (2010b) compiled chitinozoan presence/absence data for the glacial Hirnantian (equivalent to the extraordinarius and lower persculptus graptolite biozones) from localities largely within the circum-Iapetus region. These Hirnantian distribution data have been published using the new 'BugPlates' reconstruction and can therefore be considered in our analysis here without replotting.
Differences in the latitudinal position of our Sandbian localities inferred from the 'BugPlates' and the earlier PALEOMAP palaeogeographical reconstructions are shown in supplementary material: the differences are usually within our defined 58 bin of palaeolatitudinal error, with the exception of three site-clusters where the palaeogeographical error is systematically larger. Of these three areas, the large uncertainty on Avalonian localities causes most problems and these sites will not be used to pinpoint the positions of important climate boundaries in this study.
We re-plot the groups of species obtained by multivariate analyses by Vandenbroucke et al. (2009Vandenbroucke et al. ( , 2010a and called biotopes, on the new 'BugPlates' palaeogeographical reconstructions. The individual species are not shown, although the full species lists are given in the supplementary material and the relatively small differences between the published biotopes (Vandenbroucke et al. 2009(Vandenbroucke et al. , 2010a on the one hand and the biotopes in Figures S1 and S2 on the other, are listed in Table S2. The full data matrices are uploaded as supplementary Figures S3 and S4. The general pattern of Sandbian biotope distribution is little affected by the use of the 'BugPlates' palaeogeographical basemaps, compared with previously published distributions using PALEOMAP (Vandenbroucke et al. 2009(Vandenbroucke et al. , 2010a. The biotopes (e.g. C1 -C7) group into 'bioprovinces', which appear comparable to modern plankton foraminiferan provinces and, following comparison with various numerical climate models, are used to define Sandbian climate belts (exactly how this is done is discussed in full in the supplementary material). This produces a robust pattern for all available biogeographical reconstructions for the early Sandbian.

Definitions of Sandbian zooplankton provinces
We define the Sandbian chitinozoan and graptolite zooplankton biotopes as follows (and based on the 'BugPlates' reconstruction): † a Tropical province from c. 0 to 208 latitude (graptolite biotopes C1, C2, D1 -D3); In our previous papers, we chose to use 'biotope' rather than 'province' as the latter has a benthic connotation in Ordovician studies, which is linked with palaeogeographically -palaeocontintentally controlled benthic faunal units. However, we now opt to formalize usage and define planktonic 'provinces', following the practice used for modern planktonic foraminiferans (Kucera 2007). The spatial distribution of these provinces is summarized in Figure 24.4.

Sandbian palaeoceonographical boundaries
Diverse graptolite assemblages thrived at low latitudes, in warmer (sub)tropical waters. Chitinozoans, in contrast, seemed to have preferentially diversified in the relatively colder surface waters at intermediate to high latitudes (Vandenbroucke et al. 2009(Vandenbroucke et al. , 2010aFig. 24.4). The oceanographic boundary known as the Subtropical convergence, that divides essentially warm tropical water from cold polar water and marks the edge of the (sub)tropics in the modern ocean, lies at c. 358S in our new reconstruction (Vandenbroucke et al. 2009(Vandenbroucke et al. , 2010a. Equator-wards of this boundary, there are several latitudinally distinct biotopes within the Tropical (see supplementary material: graptolite biotopes C1, C2, D1 -D3) and Sub-tropical (C3, C4, D4 and D5) regions. These are non-symmetrically distributed around the Equator and may reflect regional patterns of productivity or surface currents. The fossils also map the Sandbian austral Polar Front, that is, the northernmost extension of the Polar waters, between 55 and 758S. The range of positions is due to a combination of palaeogeographical errors (c. 108) and potential multiple indicators for this oceanographic boundary. We speculate that these apparent multiple indicators of the Polar Front during the early Sandbian demonstrated by our dataset (see Fig. 24.4) could be due to temporal variation in the position of this important oceanographic boundary during the gracilis time slice, as this spans at least 3 myr.
Not all the species have a narrow latitudinal range, and only the ones that do have been used to define the biotopes and provinces. To the south of the Sub-tropical convergence (Transition zone/province), some graptolite species (defined as 'wide' on Fig. S1) had geographical ranges across the Sub-polar and Polar provinces. We have speculated earlier that these could reflect mixing and/or re-distribution by the Southern Gondwanan Current (see also Herrmann et al. 2004a;Poussart et al. 1999;Wilde 1991). The same ocean surface current system could also be responsible for the redistribution of the wider-ranging elements in the Sub-polar chitinozoan biotopes.

Late Ordovician climate
Climate proxy data derived from stable isotope analyses of biogenic materials remain controversial (e.g. see contrasting views in Shields et al. 2003;Trotter et al. 2008) and depend on assumptions about primary ocean chemistry. The value of GCMs (Herrmann et al. 2004a, b) for reconstructing Late Ordovician palaeoclimate is limited by our incomplete knowledge of prevailing boundary conditions for this interval, but the use of zooplankton distribution data can help constrain GCM outputs (Vandenbroucke et al. 2009(Vandenbroucke et al. , 2010a. Graptolite and chitinozoan surface water biotopes are latitudinally strongly differentiated through the Southern Hemisphere during the Late Ordovician. The faunal equator-to-pole gradient expressed in the combined graptolite and chitinozoan data is much steeper than would be expected for a greenhouse climate. As an analogue, the present-day global distribution of planktonic foraminiferans, representative of an icehouse world, is characterized by nine SST-controlled provinces, while in the greenhouse world of the Late Albian there are only five provinces (Hart 2007, and references therein).
The steeper early Sandbian faunal gradient is comparable to that found at present, and probably suggests a comparable SST gradient. If this interpretation is correct, then the faunal data indicate that a 'modern-type' climate and cooling towards the Hirnantian glacial maximum was already underway by the early Sandbian (see also Trotter et al. 2008;Vandenbroucke et al. 2009Vandenbroucke et al. , 2010a. This is a hypothesis that contrasts markedly with other interpretations of the 'background' climate mode during the Ordovician, historically regarded as a 'Cretaceous-style greenhouse'. The distribution of graptolite and chitinozoan provinces during the Sandbian best fits the SST simulation from the GCM model (Herrmann et al. 2004a), constrained at a pCO 2 of eight times the Pre-industrial Atmospheric Level (8 Â PAL % 2240 ppm) and returning a mean global surface temperature of 15.7 8C. These values compare relatively well with more advanced Middle Ordovician GCM model runs using the GEOCLIM model (Nardin et al. 2011).
As the climate further cooled towards the Hirnantian glacial maximum, the chitinozoan provinces indicate a steeper latitudinal temperature gradient than in the Sandbian, the Subtropical Convergence at 25-308S and the austral Polar Front at c. 408S. In this reconstruction the Polar belt had expanded, while the Sub-tropical belt may have slightly narrowed (but note our 58 error bars) and the sub-polar belt had contracted radically (Fig. 24.5). Vandenbroucke et al. (2010b) hypothesized this loss of ecospace as a possible cause of extinction within the Sub-polar province. The equator-ward migration of the Polar Front to c. 408 latitude during the Hirnantian is identical to that reported for the (boreal) Polar Front during late Cenozoic glaciations (McIntyre et al. 1972;Eynaud et al. 2009), and suggests a similar fall in mean global temperature of between 3 and 5 8C. This would equate to a fall in pCO 2 from c. 8 Â PAL to c. 5 Â PAL during the Hirnantian glaciation (Petit et al. 1999; Vandenbroucke et al. 2010b).

Conclusions
The primary control on Ordovician graptolite and chitinozoan species distribution was SST (depth probably was not a primary factor). Biogeographical province boundaries for both of these groups reflect latitudinally restricted climate belts, and can therefore be used to define the position of climate-sensitive palaeoceanographical boundaries and their movement through time. In this review, early Late Ordovician zooplankton distribution patterns (Vandenbroucke et al. 2009(Vandenbroucke et al. , 2010a have been re-evaluated using 'BugPlates'. During the Sandbian the Earth had a 'cool world' climate (sensu Royer 2006) similar to that of the present day, supporting the hypothesis that global cooling had started during the Mid Ordovician or earlier. During the Hirnantian glacial maximum, the distribution of the Sub-tropical Convergence and Polar front indicates a climate not dissimilar to that during late Cenozoic glaciations. The distribution of graptolite and chitinozoan provinces during the Late Ordovician indicates that the zooplankton responded in a very similar way to modern planktonic groups. The reconstruction of climate-oceanbiosphere interactions during the Palaeozoic is entirely possible.
We are grateful to J. Nõlvak and T. Challands, who provided chitinozoan data for the original papers. B. Rickards, A. Rushton, D. Goldman, J. Maletz, A. Snelling and M. Howe helped with the original graptolite data compilation. The PRISM project, and especially H. Dowsett, A. Haywood and U. Salzmann, are acknowledged as an inspiration for the methodology used in our paper. The initial research project was funded by the Research Foundation (FWO) -Flanders; subsequent funding to TRAV is from the Centre National de la Recherche Scientifique (action SYSTER), and the Agence Nationale de la Recherche through grant ANR-12-BS06-0014 'SeqStrat-Ice'. This is a contribution to SYSTER project 'Le climat de l'Ordovicien' (INSU/CNRS) and the IGCP project 591. The comments of two anonymous referees also improved this paper.