Quaternary International 413 (2016) 44e54
Contents lists available at ScienceDirect
Quaternary International
journal homepage: www.elsevier.com/locate/quaint
Exploring paleo food-webs in the European Early and Middle Pleistocene: A network analysis
* S. Lozano a, , A. Mateos b, J. Rodríguez b a IPHES, Institut Catala de Paleoecologia Humana i Evolucio Social, Tarragona, Spain and Area de Prehistoria, Universitat Rovira i Virgili (URV), Tarragona, Spain b National Research Center on Human Evolution (CENIEH), Paseo Sierra de Atapuerca, 3, 09002 Burgos, Spain article info abstract
Article history: Food webs are networks of feeding (trophic) interactions among species. As any other network approach, Available online 19 November 2015 research on food webs focuses its analysis on the structure of direct and indirect interactions among diverse species, rather than looking at the particularities of certain taxa. In recent times, scholars have Keywords: collected an impressive amount of empirical food-web data to study present day terrestrial and aquatic Paleo food-webs habitats. Early and middle pleistocene More restrictively, this approach has also been applied to trophic interactions represented in the fossil Macro-mammals record of extinct ecosystems. Nevertheless, to our knowledge, none of them has addressed the role Homo species played by the different Pleistocene hominin species as part of such food-webs. In this work, we aim at filling this gap by focusing on the Early and Middle Pleistocene paleo-communities in Western Eurasia. Our goal is to improve our understanding on changes experienced by large mammals' interactions during this period, and shed some light on the influence of and on Homo species of those changes. We have constructed up to 27 paleo food-webs from the archaeo-paleontological record of European assemblages, covering from the Middle Villafranchian to the Late Galerian. Only large mammals have been considered, including a couple of Homo species that are present in 8 food-webs. Then, we have developed a two-steps analysis. First, we have calculated the main structural features of all the networks, and have compared them across geographical areas, periods and cases with and without Homo species. Second, we have calculated different structural centrality measures in order to assess the relevance of Homo species in their corresponding food-webs. The obtained results show that the Pleistocene food webs under study shared basic features with modern food webs, although differences in the values of some parameters might be significant. More- over, when comparing the networks across periods, we found a marked change that could be related to the Mid-Pleistocene Revolution. Finally, our results also highlight the trophic position of hominins in the web as a central species that channeled energy fluxes. © 2015 Elsevier Ltd and INQUA. All rights reserved.
1. Introduction study present day terrestrial and aquatic habitats (Martínez, 1991; Williams and Martínez, 2000; Dunne et al., 2002; Stouffer et al., Food webs are networks of feeding (trophic) interactions among 2005; Pascual and Dunne, 2006; Brose et al., 2006). Moreover, species (Cohen et al., 1990). As any other network approach, there is an increasing literature on the construction of theoretical research on food webs focuses its analysis on the structure of direct models of food-web structure to understand dynamics of ecological and indirect interactions among diverse species, rather than look- communities like, for instance, their robustness to the extinction of ing at the particularities of certain taxa. In recent times, scholars certain species or the introduction of new ones (Allesina et al., have collected an impressive amount of empirical food-web data to 2008; Stouffer and Bascompte, 2011; Stouffer et al., 2012; Capitan et al., 2013). Beyond its application to study ecosystems in vivo, the ‘food web approach’ has also been used to study trophic relationships as * Corresponding author. E-mail addresses: slozano@iphes.cat (S. Lozano), [email protected] represented in the fossil record of extinct ecosystems. Since the (A. Mateos), [email protected] (J. Rodríguez). pioneering work by Dunne et al. (2008), which constructs and http://dx.doi.org/10.1016/j.quaint.2015.10.068 1040-6182/© 2015 Elsevier Ltd and INQUA. All rights reserved. S. Lozano et al. / Quaternary International 413 (2016) 44e54 45 analyses paleo food-webs from extraordinarily well preserved shift from small to large prey, which according to Carbone et al. Cambrian records, some authors have contributed other case (1999) is 21.5 kg. Setting the threshold at 10 kg avoids excluding studies (Maschner et al., 2009; Roopnarine and Hertog, 2010; Lotze medium-sized carnivores that occasionally include large prey in et al., 2011), and guidelines on how to deal with methodological their diets. specificity of paleo food-webs (Roopnarine, 2009). Nevertheless, to Putative trophic relationships between the species in the LFAs our knowledge, none of these works have applied food-web anal- were inferred on the basis of the information available about the ysis to the fossil record in order to address long-term ecological characteristics and behavior of the prey and predators as explained processes and, specifically, interactions of different Homo species in Rodríguez et al. (2012). Extrapolation of the behavior of recent with other animal taxa along the first steps of human evolution. The relatives (see references in Rodríguez et al., 2012), isotopic closest reference we have found, analyzes structural changes on (Palmqvist et al., 2008; García et al., 2009; Feranec et al., 2010; seed-dispersal interactions related to the late Quaternary mega- Bocherens et al., 2011), and paleontological or zooarchaeological faunal extinction in America (Pires et al., 2014). evidence was taken into account to infer the trophic relationships. The general objective of this article is to explore the potential of The evidence on the predatory behaviour and the potential prey of food-webs generated from paleontological fossil records as a valid Homotherium sp, Lynx issiodorensis, Lynx pardinus, Megantereon methodological approach to study dynamics in Pleistocene eco- cultridens, Panthera gombaszoegensis, Panthera pardus, Puma par- systems. To this end, we have chosen as a case study the late Early doides, Acinonyx pardinensis, Chasmaporthetes lunensis, Pachycro- Pleistocene and the Middle Pleistocene in Western Eurasia. A major cuta brevirostris, Crocuta crocuta, Pliocrocuta perrieri, Canis arnensis, climate and ecological event, known as the Mid-Pleistocene Revo- Canis etruscus, Canis mosbachensis, and Lycaon lycaonoides was lution (MPR), driven by variations in the orbital forcing of the discussed in detail in Rodríguez et al., 2012. Here we review the climate cycles occurred within this period (Maslin and Ridgwell, information available to infer the predatory behavior of the other 2005). The effects of the MPR on the climate system were partic- species included in the present work. ularly evident in the period from 1.0 Ma to 0.8 Ma, when a sub- The lion (Panthera leo) arrived in Europe 0.7e0.6 Ma. to occupy stantial increment on global ice volume occurred at 0.94 Ma, the the niche of a top predator, able to kill very large prey (Croitor and periodicity of the cycles changed from 41 ky to 100 ky and their Brugal, 2008). Modern lions are certainly the best analog for the amplitude increased (Head and Gibbard, 2005). These climate predatory behavior of Pleistocene P. leo. The sociability of lions changes promoted a drastic reorganization of the European eco- allow them to kill very large prey in group hunting. Their cursorial systems, that affected both the fauna and the vegetation, and was abilities and social hunting make them very efficient in open characterized by an expansion of the open environments (Suc and country, while their strong constitution makes them good ambush Popescu, 2005; Bertini et al., 2010; Croitor and Brugal, 2010; Kahlke hunters (Turner, 2009). Mean prey size of recent lions is around et al., 2011; Leroy et al., 2011; Palombo, 2014). These environmental 130 kg (correcting by age and sex of the prey), but almost half of the changes undoubtedly affected the survival opportunities of the prey weigh around 70 kg, while 40% of the kills are around 220 kg European humans because the new environments provided and the rest correspond to prey above 400 kg (Rapson and Bernard, different qualities and quantities of trophic resources (Rodríguez 2007). Megafauna species like rhinos (Brain et al., 1999) or young et al., 2012; Palombo, 2014). Within that framework, our goal is elephants are opportunistically killed and there is practically not twofold. First, to uncover possible changes on large mammals' limit for the smaller prey (Sunquist and Sunquist, 2009). Isotopic trophic relationships along this period. Second, to analyze the role studies confirm the role of Pleistocene lions as top predators, (either passive or active) played by different Pleistocene's Homo including in their diet significant amounts of megafauna species species in such changes. (Bocherens et al., 2015). A medium sized hyaenid attributed to the genus Hyaena is 2. Material and methods present in the Untermassfeld LFA and some other Middle Pleisto- cene localities. It is usually identified as Hyaena prisca, although 2.1. Data compilation and network construction Arribas and Garrido (2008) consider H. prisca a synonym of Hyaena brunnea (¼Parahyaena brunnea). Both the recent stripped hyaena A set of 27 European large mammal Local Faunal Assemblages (Hyaena hyaena) and the brown hyaena (H. brunnea) are omnivo- (LFAs) dated from the middle Villafranchian to the late Galerian was rous and extremely efficient scavengers. They also consume sig- selected from the literature (See Table 1 and Fig. 1). Five time pe- nificant amounts of vegetable matter and opportunistically kill riods, corresponding to the middle Villafranchian (2.6e1.8 Ma), late small mammals and other vertebrates (Burgener and Gusset, 2003; Villafranchian (1.8e1.2 Ma), early Galerian (1.2e0.78 Ma), middle Holekamp and Kolowski, 2009). Thus we consider than the Euro- Galerian (0.78e0.0.5 Ma), and late Galerian-early Aurelian (0.5e0.3 pean Hyaena was a scavenger, unable to kill ungulates. Ma) (Palombo, 2014) were distinguished, and local faunas were Two large social canid species were present in Europe during assigned to one of them according to biostratigraphic correlations the Middle and Late Pleistocene: Cuon alpinus and Canis lupus. The and numerical ages provided by the original sources. Only European representatives of the genus Cuon are considered by reasonably complete LFAs were included in the database. Although some authors as subspecies of the recent species Cuon alpinus, it is extremely difficult to establish strong criteria to determine living in East Asia, while other specialists consider them a different whether a LFA is reasonably complete, we established a rule of species (Brugal and Boudadi-Maligne, 2011). In any case, they were thumb based on the analisys of 1452 Pleistocene LFAs from Europe. hypercarnivorous canids closely related to the dhole (C. alpinus), We selected faunas with a number of primary and secondary which constitutes the better analog to infer their behavior. Modern consumers above the median of the 1452 LFAs. Thus, we selected a dholes are group hunters. Males weigh 15e20 kg and females number of faunal assemblages with more than seven primary 10e12 kg. They hunt medium sized and large ungulates and occa- consumer species and more than four secondary consumers. This is sionally eat carrion (Sillero-Zubiri, 2009). Preferred prey are in the a conservative criterion, based on selecting the richest LFAs on the size interval 15e50 kg and include fawns of Cervus unicolor, and confidence that they are reasonably complete, but it does not Axis axis together with adults of Muntiacus muntjak and Sus scrofa necessarily imply that poorer LFAs are incomplete. The dataset (Venkataraman et al., 1995; Andheria et al., 2007; Wang and includes only large mammals, defined as species weighing more Macdonald, 2009). The European Cuon was larger in size than the than 10 kg. This size is slightly below the threshold where predators dhole, specially the middle Pleistocene forms usually classified as 46 S. Lozano et al. / Quaternary International 413 (2016) 44e54
Cuon priscus (Brugal and Boudadi-Maligne, 2011). European dholes carrion opportunistically consumed. The species U. dolinensis is a most likely preferred prey in the 10e45 kg weight interval, but small-sized bear described on material from Atapuerca Gran Dolina cooperative hunting allowed them to effectively kill larger prey. site (García and Arsuaga, 2001) that may be present also in Species in the 45e90 kg interval, and even in the 90e180 kg in- Untermassfeld (Kahlke and Gaudzinski, 2005). A diet similar to U. terval were likely killed, specially young and physically depleted etruscus is assumed. individuals. Carrion was probably also eaten. The lineage Ursus deningeri-Ursus spelaeus represents an The diet of extant wolves is extremely variable, but ungulates evolutionary trend towards a herbivorous specialization. Isotopic become their main prey when available (Ansorge et al., 2006; data on the population of the Sima de los Huesos site suggest that Sillero-Zubiri, 2009). Pack hunting allows gray wolves to kill large the diet of U. deningeri included a large amount of vegetal matter ungulates like red deers, bisons, moose, or horses (Garrott et al., (García et al., 2009). The herbivorous behavior of the cave bear (U. 2007; Van Duyne et al., 2009). However, in Europe small un- spelaeus) was initially inferred on the basis of the morphology and gulates are the preferred prey (Okarma, 1995; Ansorge et al., 2006), wear pattern of its dentition (Kurten, 1958) and it was eventually while large ungulates seem to be the main prey in North America confirmed by ecomorphological studies (van Heteren et al., 2014), (Milakovic and Parker, 2011). Isotopic evidence from the Late and by isotopic analyses (Bocherens et al., 1994, 2014). Here we Pleistocene site of Valdegoba suggest that aurochs were part of the consider U. spealeus a strictly herbivorous species, but it is diet of Pleistocene wolves, while smaller prey like Castor fiber were admitted that U. deningeri was able to sporadically kill medium negatively selected (Feranec et al., 2010). Thus, we may assume for sized ungulates. Isotopic analyses of subfossil brown bears from the Pleistocene wolfs hunting preferences more similar to the the Alps show a mainly vegetarian diet (Bocherens et al., 2004). Northamerican than to the European recent populations. Preferred Bocherens et al. (2015) found isotopic evidence of a high con- prey size would be in the interval 90e360 kg, while prey in the sumption of mammoth meat by brown bears at the Gravettiann 10e90 kg and 360e1000 kg would be hunted as secondary prey. site of P redmostí, although they interpret that mammoths were Ursids are, generally speaking, omnivorous species, although mostly consumed as carrion. American grizzly bears (U. arctos the diet of the recent species ranges from strictly herbivorous in the horribilis) are able to kill large ungulates like moose, red deers, or giant panda (Ailuropoda melanoleuca) to hypercarnivorous in the reindeers, although meat proportion in the diet of recent euro- case of the polar bear (Ursus maritimus). Similarly, the diet of the pean brown bears is in the range 9e15% (Bocherens et al., 2004). Pleistocene species of the genus Ursus may be classified as different Concerning the Asiatic black bear (U. thibetanus) we assume for degrees of omnivory. Proportion of meat in the diet of recent bears the Pleistocene populations a behavior similar to their living varies between species and populations but, as a general rule, larger counterparts. Asiatic black bears eat large amounts of vegetable animals and more northern populations include a larger amount of food, including fruits in a large proportion, although small un- meat in their diet. Scavenging is a common practice for all bears gulates may compose a sizable proportion of their diet in some although some species like brown bears (U. arctos) are also able to areas (Garshelis, 2009). In our analysis, we considered that Pleis- kill large ungulates both fawns and adults (Garshelis, 2009). Six tocene brown bears and Asiatic black bears consumed all other species in the genus Ursus have been recorded in the set of 27 species as carrion, although they were also able to hunt small- Pleistocene LFAs: U. etruscus, U. dolinensis , U. thibetanus, U. denin- sized ungulates. geri, U. spelaeus and U. arctos. Since cannibalism and inter-specific aggression between pred- Palmqvist et al. (2008) infer for U. etruscus a diet similar to the ators are widespread behaviors (Palomares and Caro, 1999; recent U. arctos, based on their similar tooth morphology. Ursus Azevedo et al., 2010; Balme and Hunter, 2013), although not very etruscus was likely an omnivorous bear that mainly relied on frequent, it has been assumed that all carnivore species practiced vegetable food on temperate environments. Likely, ungulates in the cannibalism sporadically and that larger carnivores occasionally 10e360 kg body weight interval were sporadically hunted and killed and consumed smaller competitors.
Table 1 The 27 paleontological sites providing the empirical data from NQMDB (Neogene-Quaternary Mammals Database.
Locality Land mammal age Reference
Swanscombe Lower loam late Galerian/Aurelian (Ashton et al., 1994) Heppenloch late Galerian/Aurelian (Koenigswald and Heinrich, 1999) Atapuerca-Galería GIIb late Galerian (Rodríguez et al., 2011) Caune de l'Arago late Galerian (Alberdi et al., 1997) Atapuerca-Galería GIIa late Galerian (Rodríguez et al., 2011) Hundsheim late Galerian (Koenigswald and Heinrich, 1999) Mauer late Galerian Wagner et al. (2011) Voigtstedt middle Galerian (Koenigswald and Heinrich, 1999) Sussenborn middle Galerian (Koenigswald and Heinrich, 1999) Slivia middle Galerian (Palombo et al., 2000e2002) Atapuerca Dolina TD6 early Galerian (Rodríguez et al., 2011) Kozarnika 12 early Galerian (Sirakov et al., 2010; Martínez-Navarro et al., 2012) Untermassfeld early Galerian (Koenigswald and Heinrich, 1999) Grotte du Vallonnet III early Galerian (Echassoux, 2004) Pirro Nord late Villafranchian (Arzarello et al., 2007; Pavia et al., 2012) Ceyssaguet 1 late Villafranchian (Aouadi, 2001; Kaiser and Croitor, 2004; Tsoukala and Bonifay, 2004) Fuente Nueva-3 late Villafranchian (Martínez-Navarro et al., 2010) Venta Micena late Villafranchian (Palmqvist et al., 2010; Duval et al., 2011) Barranco Leon V BL-D late Villafranchian (Toro-Moyano et al., 2013)(Martínez-Navarro et al., 2010) Cueva Victoria late Villafranchian (Gibert, 1993; Madurell-Malapeira et al., 2014) Casa Frata late Villafranchian (Palombo et al., 2000e2002) Olivola late Villafranchian (Palombo et al., 2000e2002) Fonelas P-1 middle Villafranchian (Viseras et al., 2006; Arribas, 2008) S. Lozano et al. / Quaternary International 413 (2016) 44e54 47
Table 1 (continued )
Locality Land mammal age Reference
Puebla de Valverde middle Villafranchian (Alberdi et al., 1997) Saint-Vallier LD3 middle Villafranchian (Guerin, 2004) Gerakarou middle Villafranchian (Palombo et al., 2006) Seneze middle Villafranchian (Palombo and Valli, 2003e2004)
Fig. 1. Geographical distribution of the palaeontological sites providing the empirical data on trophic relationships. 1: Swanscombe Lower loam; 2: Heppenloch; 3: Atapuerca- Galería GIIb; 4: Caune de l'Arago; 5: Atapuerca-Galería GIIa; 6: Hundsheim; 7: Mauer; 8: Voigtstedt; 9: Sussenborn; 10: Slivia; 11: Atapuerca Dolina TD6; 12: Kozarnika 12; 13: Untermassfeld; 14: Grotte du Vallonnet III; 15: Pirro Nord; 16: Ceyssaguet 1; 17: Fuente Nueva-3; 18: Venta Micena; 19: Barranco Leon BL-D; 20: Cueva Victoria; 21: Casa Frata; 22: Olivola; 23: Fonelas P-1; 24: Puebla de Valverde; 25: Saint-Vallier LD3; 26: Gerakarou; 27: Seneze.
The concrete locations of the paleontological sites providing the C (L/S2): Connectance or density of trophic relationships in the data are shown in Fig. 1. network. It is calculated as the total number of connections These data on trophic relations were used to generate one food- identified in the network over the total possible existing ones. web per site. Specifically, food-webs were constructed as (directed) Consequently, its possible values are bounded between 0 and 1. graphs, where species were represented by vertices and trophic L/S: Average number of connections per species (also referred to relationships corresponded to arcs pointing to predators (see Fig. 2 as average degree in network analysis). This indicator provides for clarifications). Notice that our nodes represent original and not information on specialization. The more specialized the carni- 'trophic species', as is usually done in food web analysis (Dunne, vores in the network (or the fewer predators prey herbivores), 2009). “Trophic species” are equivalent trophic positions in a the lower the value of L/S. network that might be occupied by several original species, pro- Path: Average path (geodesic) length or, in other words, the vided that they share the same predators and prey. The reason for average number of steps along the shortest paths for all possible such a choice is that we aim at focusing our study on the role of a pairs of network nodes It can be understood as a global measure specific set of species, and in particular Homo. of trophic positions averaged across species in the web Both the construction of these paleo food webs and their analysis (Williams and Martínez, 2000; Capitan et al., 2013). (Subsections 2.2 And 2.3) were conducted using Python and the Path_r: Average value of Path for n samples of an Erdos-Renyi€ IGRAPH library of R statistical package (See http://igraph.org/r/). random graph (Bollobas, 2001) equivalent to the empirical food- web under study. Erdos-Renyi€ graphs are regular random 2.2. General structural analysis graphs where vertices are linked with a given probability p. This regularity makes it easy to mathematically characterize this sort Once the networks were constructed, we conducted a basic of graphs and, as a consequence, they are commonly used as null structural analysis in order to get a general overview of the models to evaluate empirical features like the Path value. If Path resulting paleo food-webs, and make a comparison among them and Path_r are significantly different, we consider that this (across different periods and geographical regions) and with feature contains relevant particular information of the food-web modern ones. First, we calculated some simple structural measures (usually related to functional aspects or formation process). typically applied to this sort of networks (Dunne, 2009), namely: CI (Clustering coefficient): Density of triads/triangles in the network (e.g. Ae > Be > Ce > A). Its value is bounded between S and L: Respectively the total number of species and relation- 0 and 1. ships composing a food-web. 48 S. Lozano et al. / Quaternary International 413 (2016) 44e54
Fig. 2. Three examples of trophic relations as represented in a food web. Left: A feeds B (or, in other words, B predates A); Centre: A and B feed (and predate) each other; Right: A feeds B, which also practices cannibalism.
Fig. 3. Two examples of paleo food-web visualization corresponding to Venta Micena (left) and Atapuerca Galeria GIIa (right). Both figures were created using the Gephi network analysis tool (http://gephi.github.io/).