Journal of Mammalian Evolution

, Volume 14, Issue 1, pp 1–35

Castorid Phylogenetics: Implications for the Evolution of Swimming and Tree-Exploitation in Beavers

Authors

    • Earth Sciences/PaleobiologyCanadian Museum of Nature
Original Paper

DOI: 10.1007/s10914-006-9017-3

Cite this article as:
Rybczynski, N. J Mammal Evol (2007) 14: 1. doi:10.1007/s10914-006-9017-3

Abstract

Beavers (Castoridae) are semiaquatic rodents that modify forest and aquatic habitats by exploiting trees as a source of food and building material. The capacity of beavers to transform habitats has attracted interest from a variety of researchers, including ecologists, geomorphologists and evolutionary biologists. This study uses morphological and behavioral evidence from the fossil record to investigate the evolutionary history of tree-exploitation and swimming in beavers. The findings suggest that both behaviors appeared within a single castorid lineage by the beginning of the Miocene, roughly 24 million years ago. Biogeographic results support the hypothesis that tree-exploitation evolved at high latitudes, possibly influenced by the development of hard winters.

Keywords

BeaverBehavioral evolutionBiogeographyCastorFossilPaleontologyPhylogeneticsWoodcutting

Introduction

The extant beaver, Castor (Castoridae), is a Holarctic rodent that harvests trees for food and construction (ex. lodge and dam building), and in so doing influences local geomorphology, hydrology, freshwater community structure, plant succession and species richness (Butler, 1995, p. 215–222; Wright and Jones, 2002; Müller-Schwarze and Sun, 2003). Of particular interest is whether Castor foraging and construction activities have a selective effect on the evolution of associated populations (see Jones et al., 1994; Odling-Smee et al., 2003). For example, there is some evidence to suggest that tree-exploitation by Castor may have favored the evolution of chemical defenses in poplar trees (Basey et al., 1990; Basey and Jenkins, 1993). Also, the evolution of aquatic taxa, such as fish and amphibians, may be affected by the presence of beaver ponds (Basey and Jenkins, 1993; Quail, 2001; Pollock et al., 2004; Ray et al., 2004). Unfortunately, in order to predict how profound or pervasive the effects of habitat modification by beavers might be, it is important to first understand the evolutionary history of habitat modification by beavers. The temporal scale of the engineering effects on a landscape is of particular importance. If beavers have been “engineering” habitats for millions of years we would expect their ecological and landscape impact to be greater than if the engineering behaviors were present for only a few thousand years. Toward this end, the present study uses phylogenetic methods and fossil behavioral evidence to investigate the evolutionary origins of swimming and tree-exploitation in castorids.

Castoridae is a family of Holarctic, herbivorous rodents that originated by the latest Eocene of North America and is represented in the fossil record by roughly 30 genera (Kardong et al., 1997; Korth, 2002; Korth and Rybczynski, 2003). The group is morphologically and behaviorally diverse and includes a lineage of small-bodied burrowing specialists (<1.5 kg, Reynolds, 2002), as well as large-bodied semiaquatic forms, most notably the giant, Ice-Age, Castoroides (60–100 kg, Reynolds, 2002).

Evidence of beaver cut-wood is relatively common in the fossil record (Harington, 1977; Aalto et al., 1989; Muhs et al., 1997; Matheus et al., 2000). Usually the fossil woodcuttings are Pleistocene in age and attributable to the modern beaver, Castor. There have been some reports of Castoroides woodcuttings (e.g. Hillerud, 1975, 1976); unfortunately, these claims have not been substantiated. Rather, the first definitive evidence of tree-exploitation by a genus other than Castor has been associated with Dipoides (Harington, 1996; Harington, 1990; Tedford and Harington, 2003). The remains of Dipoides in association with over 100 specimens of cut wood have been recovered from the Beaver Pond locality (Fig. 1), an early Pliocene deposit in the Canadian High Arctic (78°33′ N, 82°25′ W). The cut marks (Fig. 2) occur on branches, trunks and roots, and appear to be the result of both feeding and harvesting behavior (Rybczynski, 2003). Some of fossil cut-sticks were found intertwined near the base of the deposit, suggesting the remains of a “nest” structure, such as a lodge or dam (Tedford and Harington, 2003).
https://static-content.springer.com/image/art%3A10.1007%2Fs10914-006-9017-3/MediaObjects/10914_2006_9017_Fig1_HTML.gif
Fig. 1

Map of Arctic showing location of Beaver Pond locality, indicated by a star. The Beaver Pond deposit is 4–5 million years old and preserves the remains of the extinct castorid, Dipoides sp., in association with an assemblage of cut wood

https://static-content.springer.com/image/art%3A10.1007%2Fs10914-006-9017-3/MediaObjects/10914_2006_9017_Fig2_HTML.gif
Fig. 2

Example of cut marks on wood (a) compared to lower incisor shape of Dipoides sp. (b, c). Left lower incisor of Dipoides sp. (CMN 51759) shown in anterior view (b) and 3/4 lateral view (c). Light grey area at the tip of the incisor has been reconstructed. As in other rodents, the upper incisors were used for anchoring the head, while the lower incisors were used for cutting (20). Abbreviation: CMN, Canadian Museum of Nature. Scale bar, 1 cm

The phylogenetic relationship between Castor and Dipoides has important implications for our understanding of the origins of tree-exploitation. If Castor and Dipoides are members of ecologically similar, sister lineages (see Korth, 2002), it would be reasonable to hypothesize that their most recent common ancestor also exploited trees. However, some classifications (e.g. Xu, 1995; McKenna and Bell, 1997) suggest that Castor, but not Dipoides, evolved from a burrowing specialist, adapted to open-habitats (see Martin, 1987; Korth and Rybczynski, 2003), implying that Castor’s ancestor was not specialized for exploiting trees. Such a phylogenetic relationship would suggest that tree-exploitation arose multiple times within Castoridae. To evaluate these conflicting hypotheses this study optimizes behavioral characters (e.g., swimming, tree-exploitation) onto a morphologically-based castorid phylogeny. The resulting hypothesis of behavioral evolution is discussed in the context of biogeographic and climate evidence.

Materials and methods

Phylogenetic relationships within Castoridae were estimated using a matrix of 38 taxa and 88 morphological characters, including 66 cranial and 22 postcranial characters. Characters and taxa were chosen with the aim of recovering major patterns of diversification, rather than lower-level phylogenetic affinities. Some of the characters were modified from previous phylogenetic studies (Stirton, 1935; Wahlert, 1972, 1977; Xu, 1995; Korth, 2002) but 30 of the cranial characters and all 22 postcranial characters are new. Fifteen characters are quantitative. Quantitative character states were defined along natural breaks in the sample, but if natural breaks were absent, states were delimited by dividing the sample into even ranges. Outgroup taxa include the castorid fossil sister-taxon, Eutypomys (Korth, 2002), and the morphologically conservative, early rodentiaforms, Paramys copei, P. delicatus and Pseudotomus robustus (see Appendix A). The ingroup includes representatives of all major lineages, with a particular emphasis on “primitive” taxa. All taxa used in the analysis were at least 30% complete for the coded characters. Character definitions and the data matrix appear in Appendix B and C, respectively.

Parsimony analysis was conducted using the PC version of PAUP* 4.0b10 (Swofford, 1998), with Paramys copei, P. delicatus, Pseudotomus robustus and Eutypomys thomsoni designated as outgroup taxa. In the initial analysis all characters were unordered and equally weighted. The heuristic search employed “tree-bisection and reconnection” (TBR) branch-swapping starting from ten random step-wise addition sequence replicates, with ten trees held at each step. Characters were then reweighted using the rescaled consistency index (Farris, 1969) and the analysis was rerun, resulting in a single most parsimonious tree.

Bootstrapping was used to assess the uncertainty of the phylogenetic estimates. The dataset was not amenable to standard bootstrap methods (i.e., involving heuristic searches), so bootstrap values were estimated from a fast heurstic search (i.e., “SEARCH=FASTEP”) in PAUP, using 100,000 replicates. Characters were weighted, as described above, and the characters were sampled with equal probability by specifying “WTS = SIMPLE”.

A hypothesis of behavioral evolution was generated by optimizing behavioral character states (Table 1) onto the reweighted tree, using MacClade, version 3.01 (Maddison and Maddison, 1992). Two behavioral “characters” were considered. The first character refers to locomotor behavior, and considers taxa to be either semiaquatic or terrestrial. The second behavioral character contrasts tree-exploiting, with a preference for open-habitats. Taxa that exploit trees (i.e., Dipoides and Castor) are associated with treed habitats, including forested, river-corridor habitats. Open habitats, on the other hand, being generally treeless, are assumed to preclude tree-exploitation. Seven castorids are considered to be associated with open-habitats (Table 1). These castorids are fossorial specialists from the Harrison formation, a deposit interpreted paleoecologically to have been a dry, open, upland habitat (Martin, 1987).
Table 1

Behavioral character statesa

 

Earliest occurrence

  

Taxon

in the fossil record

Behavior

References

Agnotocastor coloradensis

Earliest Oligocene

  

Agnotocastor praetereadens

Early Oligocene

  

Agnotocastor sp

Early Oligocene?

  

Anchitheriomys fluminis

Early Miocene

  

Anchitheriomys tungerensis

Middle Miocene

  

Anchitheriomys wiedenmanni

Middle Miocene

  

Anchitheriomys sp

Miocene

  

Capacikala sp.1

Late Oligocene

  

Capacikala sp.2

Late Oligocene

  

Castor fiber

Late Miocene

Woodcutting/Semiaquatic/Burrowing

(Novak, 1987)

Castoroides ohioensis

Earliest Pleistocene

Semiaquatic

(Moore, 1890)

Dipoides sp

Early Pliocene

Woodcutting

(Rybczynski, 2003)

Dipoides smithi

Late Miocene

  

Dipoides stirtoni

Late Miocene

  

Dipoides tanneri

Late Miocene

  

Dipoides c.f

Late Miocene

  

Dipoides tanneri

   

Eucastor tortus

Middle Miocene

  

Euhapsis breugerorum

Early Miocene

Terrestrial/Open-habitat Fossorial

(Martin, 1987)

Euhapsis ellicotae

Early Miocene

Terrestrial/Open-habitat Fossorial

(Martin, 1987)

Euroxenomys minutum

Early Miocene

  

*Eutypomys thomsoni

Late Eocene

Terrestrial

(Korth, 1994)

Fossorcastor greeni

Early Miocene

Terrestrial/Open-habitat Fossorial

(Martin, 1987)

Hystricops venustus

Middle Miocene

  

Migmacastor procumbodens

earliest Miocene

Terrestrial

(Korth and Rybczynski, 2003)

Palaeocastor fossor

Late Oligocene

Terrestrial/Open-habitat Fossorial

(Martin, 1987)

Palaeocastor magnus

Early Miocene

Terrestrial/ Open-habitat Fossorial

(Martin, 1987)

Palaeocastor c. f. nebrascensis

Middle Oligocene

  

Palaeocastor peninsulatus

Late Oligocene

  

Palaeocastor c.f. simplicidens

Early Miocene

  

Palaeocastor wahlerti

Early Oligocene

  

Palaeocastor sp

Early Miocene

  

*Paramys copei

Early Eocene

Terrestrial

(Korth, 1994)

*Paramys delicatus

Middle Eocene

Terrestrial

(Korth, 1994)

Priusaulax browni

Earliest Miocene

  

Procastoroides idahoensis

Early Pliocene

Semiaquatic

(Shotwell, 1970)

Pseudopalaeocastor barbouri

Late Oligocene

Terrestrial/Open-habitat Fossorial

(Martin, 1987)

*Pseudotomus robustus

Middle Eocene

Terrestrial

 

Sinocastor anderssoni

Late Miocene

  

Steneofiber depereti

Early Miocene

  

Steneofiber eseri

Latest Oligocene

Swimming/Burrowing

(Hugueney and Escuillié, 1995, 1996)

Trogontherium cuvieri

Latest Pliocene

Semiaquatic

(Schreuder, 1929, 1951)

New Taxon A

Early Miocene

Terrestrial/Open-habitat Fossorial

 

aTaxa marked with an asterisk are outgroup to Castoridae. “Woodcutting” refers to tree/branch harvesting.

The fossorial taxa tend to exhibit robust forelimbs, large front claws, a shortened neck, and a reduced tail (Korth, 2002, 1994; Martin, 1987; Peterson, 1905). Included among these are Pseudopalaeocastor barbouri, Palaeocastor fossor, and P. magnus, which have been found associated with fossil burrows. The exceptionally well-preserved burrows of P. fossor, and P. magnus are notable because the sides of the burrows show tooth marks and claw marks, indicating that both species excavated using their forelimbs and incisors (see Martin and Bennett, 1977). Rodents that dig with their incisors (i.e., “tooth-diggers”) tend to exhibit certain craniodental specializations, such as procumbent, elongated incisors. Based on the presence of such traits, other castorids considered to tooth diggers include: Euhapsis breugerorum, E. ellicotae, Fossorcastor greeni, Palaeocastor fossor, P. magnus (Martin, 1987) and Migmacastor (Korth and Rybczynski, 2003).

Table 1 shows eleven terrestrial taxa, and four semiaquatic taxa. The terrestrial taxa include the aforementioned fossorial forms, and also the outgroup lineages, Paramys, Pseduotomus and Eutypomys (Korth, 1994). The taxa considered to be semiaquatic are Castoroides, Procastoroides, Steneofiber eseri and Trogontherium. These forms tend to exhibit a shortened femur, enlarged hind-feet, a foot morphology consistent with the presence of webbing, and specialized caudal vertebrae (Moore, 1890; Schreuder, 1929, 1951; Shotwell, 1970; Hugueney and Escuillié, 1995; Hugueney, 1999). The presence of a specialized grooming claw, seen in multiple semiaquatic taxa, including Steneofiber eseri and Procastoroides (Shotwell, 1970; Hugueney and Escuillié, 1995), might also be linked with swimming. In Castor this claw is considered to function in maintaining a “non-wettable fur”, important for buoyancy (e.g., Hugueney and Escuillié, 1995).

Another castorid considered to have been semiaquatic is the “primitive” Agnotocastor. It seems this taxon was originally suggested to have been semiaquatic because its remains were often recovered from pond deposits (Martin, 1987). Further morphological study is required, but preliminary observations of the postcranial material from Agnotocastor coloradensis (KUVP 10986) did not initially reveal any morphological correlates for swimming (pers. obs.). For the purposes of this analysis the locomotor behavior of Agnotocastor is treated as “unknown”.

Results and discussion

Phylogenetic analysis

The single most parsimonious tree (Fig. 3; Table 2) resulting from successive reweighting of characters recovered two clades along with a single monotypic lineage, represented by Agnotocastor coloradensis. Agnotocastor is paraphyletic, with Agnotocastor sp. and A. praeteraedens appearing as the most basal lineages within the two major clades (Fig. 3). The analysis recovered three groups that have been previously recognized, namely Palaeocastorinae, and Castoroidinae + Castorinae (see Korth, 2002).
https://static-content.springer.com/image/art%3A10.1007%2Fs10914-006-9017-3/MediaObjects/10914_2006_9017_Fig3_HTML.gif
Fig. 3

Single most parsimonious tree resulting from successive reweighting of a morphological character data matrix. The analysis recovered three clades that correspond to the classification of Korth (2002) as follows: Castorinae + Castoroidinae and Palaeocastorinae. The tree length is 96.5, consistency index 0.71, and rescaled consistency index is 0.57. Values shown at nodes are bootstrap support values. Only values greater than 50% are shown

Different from previous studies is the position of Anchitheriomys and Hystricops, both of which have been the subject of significant phylogenetic disagreement (see Xu, 1995; McKenna and Bell, 1997; Koenigswald et al., 2001; Korth, 2002). Most workers have suggested that Anchitheriomys is a member of a lineage that diverged prior to the origination of Castoroidinae + Castorinae, and Palaeocastorinae (but see Xu (1995) for a different view). This analysis found Anchitheriomys to be a sister lineage of the Castoroidinae + Castorinae, a relationship supported by four unambiguous characters.
Table 2

List of unambiguous apomorphies for major clades and groupsa

Group

Character

 

Character-state change

Castoridae

2

Rostrum cross-sectional shape

0 to 2

 

4

Nasal width

0 to 1

 

47

Upper incisor cross-sectional shape

0 to 1

 

54

Lower incisor cross-sectional shape

0 to 1

 

56

Shape of mandible anteriorly

0 to 1

 

78

Shape of calcaneoastragalus facet

0 to 1

“Burrowing clade”

15

Jugal shape

1 to 0

“Burrowing clade,” excluding “Agnotocastor” sp

17

Maxilla in palatal view

0 to 2

49

Upper third premolar

0 to 1

New Taxon A + Euhapsis + Fossorcastor + Pseudopalaeocastor + Palaeocastor fossor + P. magnus

11

Origin of superficial masseter

3 to 2

38

Mandibular glenoid

0 to 1

57

Digastric eminence

1 to 2

“Semiaquatic clade”

5

Nasal shape

0 to 1

 

20

Posterior maxillary foramen/notch

1 to 0

 

41

Auditory bulla, shape of medial process

0 to 1

 

45

Squamoso-mastoid foramen

0 to 1

“Semiaquatic Clade” excluding Agnotocastor praeteraedens

17

Maxilla in palatal view

0 to 1

39

Temporal foramina

0 to 1

49

Upper third premolar

0 to 1

58

Position of digastric process

0 to 1

“Semiaquatic Clade” excluding Agnotocastor praeteraedens and Anchitheriomys

23

Position of posterior palatine foramen

0 to 1

36

Sagittal crest shape

0 to 2

52

Molar shape

2 to 3

Castor + Steneofiber clade

10

Infraorbital foramen shape

0 to 3

20

Posterior maxillary foramen

0 to 1

 

55

Lower incisor wear

0 to 1

 

71

Shape of third metacarpal

0 to 1

 

84

Size of third metacarpal

1 to 2

(Trogontherium+Euroxenomys)+ (Castoroides+Procastoroides)

2

Rostrum cross-sectional shape

2 to 1

11

Origin of superficial masseter

3 to 4

14

Shape of bulla

1 to 0

51

Length of third molar

0 to 1

aClades and groups follow usage in text.

The phylogenetic placement of Hystricops has also been extremely variable. Some authors have considered it to be most closely related to Agnotocastor (Korth, 2002), whereas others have proposed a close relationship with Castor (Xu, 1995; McKenna and Bell, 1997). This analysis placed Hystricops in a clade with Steneofiber and Castor. The Steneofiber-Castor clade is supported by five unambiguous characters.

One difference between these results and previous studies is the position of Migmacastor. Previously, Migmacastor was considered to be outside Palaeocatorini and Castoroidinae + Castorinae (Korth and Rybczynski, 2003). Here, Migmacastor and Agnotocastor sp. appear to be successive sister-taxa to all other taxa within the Palaeocastorinae clade, a relationship supported by two characters (see Table 2).

An unexpected result of this analysis is the position of Trogontherium + Euroxenomys as a derived member within Castoroidini, and sister to Procastoroides + Castoroides. This position for Trogontherium + Euroxenomys is suspect, at least from the perspective of dental evolution. The cheek teeth of Trogontherium and Euroxenomys are closed rooted, and relatively low crowned, whereas the teeth of the taxa bracketing Trogontherium + Euroxenomys are open-rooted (i.e., evergrowing). Thus, the phylogeny in Fig. 3 implies that Trogontherium + Euroxenomys evolved low-crowned cheek teeth from an ancestor with high-crowned teeth. There are multiple examples in mammalian evolution of evolutionary transformation within lineages from low to high crowned teeth, but there seems to be no example of the reverse. Perhaps Trogontherium + Euroxenomys is exceptional. Alternatively, the placement of Trogontherium + Euroxenomys in this analysis is mistaken. The latter seems most likely considering that many of the morphological similarities linking Trogontherium + Euroxenomys and Procastoroides + Castoroides (see Table 2) may be the result of parallel/convergent evolution associated with large body size. Procastoroides, Castoroides, and Trogontherium are the largest of all castorids. Euroxenomys, on the other hand, is a small bodied form, that appears primitive relative to Trogontherium (Korth, 2002). To evaluate the influence of Trogontherium on the phylogenetic reconstruction the analysis was rerun (same methods as above) with Trogontherium excluded. In the resulting cladogram Euroxenomys appeared outside both major clades (tree not shown). This dramatic shift in phylogenetic position suggests that the Trogontherium + Euroxenomys lineage is more basal than is implied by the Fig. 3 phylogeny. Given that Euroxenomys appears generally relatively primitive, compared to Trogontherium, it should be more useful in reconstructing higher-level phylogenetic affinities. Unfortunately, Euroxenomys is a very poorly understood taxon, with 70% of its characters coded as unknown. A more robust estimate of the phylogenetic position of Trogontherium + Euroxenomys will require more characters and/or taxa.

Behavioral evolution

Figure 4 shows hypotheses of behavioral evolution, reconstructed by optimizing behavioral characters (Table 1) onto the morphological cladogram shown in Fig. 3. Only the most parsimonious optimizations are shown. Figure 4A shows the optimization of terrestrial and semiaquatic states. Taxa specialized for aquatic locomotion are confined to the Castoroidinae + Castorinae clade. The Migmacastor + Palaeocastorinae clade appears to have been entirely terrestrial. Given that the outgroup taxa also are considered terrestrial, it would appear that castorids were primitively terrestrial.
https://static-content.springer.com/image/art%3A10.1007%2Fs10914-006-9017-3/MediaObjects/10914_2006_9017_Fig4a_HTML.gifhttps://static-content.springer.com/image/art%3A10.1007%2Fs10914-006-9017-3/MediaObjects/10914_2006_9017_Fig4b_HTML.gif
Fig. 4

Evolutionary history of semiaquatic versus terrestrial habits (A), and woodcutting versus open-habitat fossoriality (B) in Castoridae. Behavioral evolution is reconstructed by optimizing characters onto phylogeny shown in Fig. 3. Taxa shown in boxes are associated with known behaviors (see Table 1). Behaviors of taxa without boxes are inferred based on their phylogenetic position. Only unequivocal optimizations are shown, taxa whose behavior could not be reconstructed are shown with grey branches. In ‘A” white boxes and branches indicate semiaquatic taxa, whereas terrestrial taxa are indicated by black boxes and branches. In “B” white and black branches/boxes are associated with woodcutting and open-habitat fossoriality, respectively

Figure 4B shows the most parsimonious interpretation for the evolution of specialized tree-exploitation (i.e., tree/branch harvesting) and fossoriality. All the fossorial specialists identified in this analysis are confined to a single major clade and, as is typical among subterranean mammals today (Nevo, 1979), they are associated with arid, open habitats (Martin and Bennett, 1977; Martin, 1987; Korth and Rybczynski, 2003). Within the fossorial group (i.e., Migmacastor + Palaeocastorinae) the most morphologically specialized taxa exhibit craniodental traits that seem associated with tooth-digging (see discussion in Korth and Rybczynski, 2003). Tooth-digging appears to have evolved at least twice in the fossorial clade: once in a group including Palaeocastor fossor and once in a lineage represented by Migmacastor (Korth and Rybczynski, 2003). These two tooth-digging lineages are separated by taxa which appear relatively unspecialized with respect to tooth-digging.

Specialized fossoriality is confined to the Migmacastor + Palaeocastorinae clade, yet burrowing, as a more generalized behavior, characterizes at least some of the semiaquatic castorids as well. Castor, for example, is not generally considered a fossorial specialist, but it is known to dig bank-burrows and canals, the latter that can be over one hundred meters long (Butler, 1995). Steneofiber also seems to have been a burrower. A fresh water limestone deposit of Montaigu-le-Blin in France yielded the remains of a Steneofiber family that apparently died within a fossil burrow (Hugueney and Escuillié, 1995). Given that burrowing behavior is characteristic of both the semiaquatic and burrowing clades, it seems likely that burrowing behavior arose primitively within Castoridae, if not earlier.

Castor and Dipoides, the only taxa known to exhibit tree/branch harvesting, are members of the semiaquatic clade. They form a phylogenetic bracket around most of the semiaquatic taxa (keeping in mind that the phylogenetic position Trogontherium + Euroxenomys is dubious, see above). If tree-exploitation evolved once, most if not all of the taxa within the semiaquatic clade must have shared a common ancestor that was a tree-exploiting specialist. The oldest taxon in this tree-exploiting group is Steneofiber eseri, which appears in the fossil record in the earliest Miocene (Fig. 5). Other fossil evidence, too fragmentary to be included in this analysis, suggests that the Steneofiber lineage can be traced into the Oligocene (Hugueney, 1999).
https://static-content.springer.com/image/art%3A10.1007%2Fs10914-006-9017-3/MediaObjects/10914_2006_9017_Fig5_HTML.gif
Fig. 5

Temporal and geographical ranges of representative castorid taxa. The geographical range of the Burrowing clade is restricted to North America whereas the Semiaquatic Clade is Holarctic. The common ancestor of the clade which gave rise to the tree-exploiting taxa, Castor and “Dipoides”, originated by 25 million years ago. Temporal and geographical data are derived from Xu (1995), Hugueney (1999), Korth (2001, 2002). Abbreviations: Eoc, Eocene; Plio, Pliocene; P., Pleistocene + Holocene

Possibly, the earliest woodcutting castorid was a semiaquatic, burrowing form (see Wilsson, 1971) whose diet included the low, woody vegetation bordering streams. Alternatively, tree-exploitation may have arisen earlier, perhaps in a late Eocene terrestrial ancestor. A modern analogue for a more terrestrial ancestral form might be Aplodontia, the mountain beaver, a non-hibernating, burrowing rodent that climbs shrubs and small trees to cut off twigs and branches (Nowak, 1999). If tree-exploitation originated in the castorid stem-lineage, then supposedly this behavior would have been lost in the fossorial clade, whose members were mostly open-habitat specialists, with limited opportunity to exploit woody vegetation.

Conclusions and environmental considerations

The capacity of the modern beaver, Castor, to engineer forest and wetland habitats is underlain by a complex of behaviors, including tree/branch harvesting (i.e., woodcutting) and swimming. Previous work (see Table 1) indicates that swimming and woodcutting are shared with some fossil taxa, but not others. Burrowing castorids associated wih open-habitats, were neither specialized for woodcutting, nor specialized for swimming. By mapping these various behaviors onto a morphologically based phylogeny, this study provides a framework for inferring the evolutionary history of swimming and woodcutting in Castoridae. The results suggest that swimming and woodcutting evolved once within Castoridae. The hypothesized common ancestor lived at least 24 million years ago (see Figs. 4 and 5), was probably also a burrower, and survived to give rise to a diverse clade spanning over 60-fold range in body size, and at least eight genera (including the extant Castor).

Unlike the fossorial clade, which was confined to North America, the biogeography of the semiaquatic/woodcutting clade implies frequent dispersal between Eurasia and North America (Fig. 5). Faunal interchange occurred via the Beringian isthmus, an Arctic “land-bridge” that existed through most of the Cenozoic (Gladenkov et al., 2002). Located at Arctic latitudes, the Beringian isthmus, would have been highly susceptible to climate change and would have functioned as a variable filter to mammalian dispersal (Beard and Dawson, 1999). Dispersal was facilitated by the warm global conditions of the early Cenozoic (Beard and Dawson, 1999). In contrast, cooler conditions in the later Cenozoic (Zachos et al., 2001) would have made it difficult for cold-sensitive to exploit high latitudes. As fossil evidence from Devon Island (75°22′ N) suggests, the High Arctic had succumbed to hard winters, including snow accumulation and lake-freezing by the Early Miocene (Whitlock and Dawson, 1990). Hard winters do not appear to have restricted populations of semiaquatic castorids from dispersing through Arctic regions. One factor underlying the success of the semiaquatic castorids at high latitudes might have been their ability to to derive food and shelter from trees.

Forests have dominated Arctic regions for most of the Cenozoic (Mathews and Ovenden, 1990; Whitlock and Dawson, 1990; McIver and Basinger, 1999; Mathews and Fyles, 2000; Williams et al., 2003) and probably throughout this time were an important resource for Arctic Herbivores. For many extant, temperate mammalian herbivores, including Castor, woody vegetation is a critically important winter food (Swihart and Bryant, 2001). Modern Castor survives harsh subfreezing winters by constructing a cache of branches in the fall. The cache is stored underwater, fastened to the pond’s substrate, and weighted down with a “cap” of less-preferred branches (Novak, 1987). Caching is initiated by the first sign of frost, and is most well-developed in northern populations of Castor. It has been proposed that caching coevolved with the development of hard winters and the freezing over of terrestrial waters (Aleksiuk, 1970; Lancia et al., 1982).

Nest structures, such as dams and lodges, are also linked with winter survivorship. Dam building, although probably favored by multiple selective factors (Müller-Schwarze and Sun, 2003), provides a deep-water location for under-ice food caching (Wilsson, 1971; Müller-Schwarze and Sun, 2003). Also, a study of Ondatra (muskrat) lodges and burrows found that in the winter, lodges were significantly warmer than burrows (MacArthur and Aleksiuk, 1979). So for population(s) of semiaquatic castorids the Arctic would have presented a selective regime favoring the construction of food caches, lodges, and dams. Although the degree to which “engineering” behaviors were developed in other extinct castorid lineages remains to be examined, it is hypothesized that high-latitude habitats may have played a key role in the evolution of the “engineering” traits seen in Castor today.

Acknowledgments

I am enormously grateful to K. K. Smith and V. L. Roth (Duke University) for discussion and reading earlier drafts. I also thank A. Ballantyne, P. Baker, M. Cartmill, W. W. Hylander, D. Schmitt (Duke University), W. W. Korth (Rochester Institute of Vertebrate Paleontology, New York), W. McLellan (University of North Carolina, Wilmington) and C. R. Harington (Canadian Museum of Nature) for feedback during the preparation of this manuscript. I am indebted to the numerous museums who provided both kind hospitality and access to specimens. In particular I am grateful (in alphabetical order by institution) to: J. P. Alexander, R. Tedford, J. Meng (American Museum of Natural History, New York), K. Shepherd and M. Feuerstack (Canadian Museum of Nature, Ottawa), M. R. Dawson, A. Tabrum (Carnegie, Pittsburgh, USA), J. Agusti (Institut de Paleontologia “Miquel Crusafont”, Sabadell, Spain), D. Heinrich, R. Schoch (Museum für Naturkunde, Humboldt University, Berlin), P. Tassy, X. Filoreau (Museum National D’Histoire Naturelle, Paris, France), R.W. Purdy, D. Levin (National Museum of National History, Washington, DC), B. Engesser and staff (Natural History Museum, Basel, Switzerland), T. Engel (Naturhistorisches Museum Mainz), K. Heissig (Paläontologisches Museum München, Germany), G. Storch and T. Dahlmann (Senkenberg-Museum, Frankfurt, Germany), M. Hugueney (Université Claude Bernard Lyon 1, France), L. Martin, D. A. Burnham, K. Gobetz, and D. Maio (University of Kansas, Lawrence), M. Voorhies and G. Corner (University of Nebraska, Lincoln). This research was supported by funding from an Aleanne Webb Dissertation Improvement Grant (Duke University), Duke Travel Grant, Sigma-Xi Travel Grant, Polar Continental Shelf Project, and NSF Dissertation Improvement Grant IBN-0073119.

Copyright information

© Springer Science+Business Media, Inc. 2006