Novack-Gottshall, P.M. 2007. Using a theoretical ecospace to quantify the ecological diversity of Paleozoic and modern marine biotas

—The process of evolution hinders our ability to make large-scale ecological comparisons—such as those encompassing marine biotas spanning the Phanerozoic—because the compared entities are taxonomically and morphologically dissimilar. One solution is to focus instead on life habits, which are repeatedly discovered by taxa because of convergence. Such an approach is applied to a comparison of the ecological diversity of Paleozoic (Cambrian–Devonian) and modern marine biotas from deep-subtidal, soft-substrate habitats. Ecological diversity (richness and disparity) is operationalized by using a standardized ecospace framework that can be applied equally to extant and extinct organisms and is logically independent of taxonomy. Because individual states in the framework are chosen a priori and not customized for particular taxa, the framework fulfills the requirements of a universal theoretical ecospace. Unique ecological life habits can be recognized as each discrete, n-dimensional combination of character states in the framework. Although the basic unit of analysis remains the organism, the framework can be applied to other entities—species, clades, or multispecies assemblages—for the study of comparative paleoecology and ecology. Because the framework is quantifiable, it is amenable to analytical techniques used for morphological disparity. Using these methods, I demonstrate that the composite Paleozoic biota is approximately as rich in life habits as the sampled modern biota, but that the life habits in the modern biota are significantly more disparate than those in the Paleozoic; these results are robust to taphonomic standardization. Despite broadly similar distributions of life habits revealed by multivariate ordination, the modern biota is composed of life habits that are significantly enriched, among others, in mobility, infaunality, carnivory, and exploitation of other organisms (or structures) for occupation of microhabitats. Philip M. Novack-Gottshall. Department of Biology, Duke University, Box 90338, Durham, North Carolina 27708-0338 Present address: Department of Geosciences, University of West Georgia, Carrollton, Georgia 30118-3100. E-mail: [email protected] Accepted: 28 January 2007 Ecological communities, however, do exist, but what are linked in them by biotic factors are not the faunistic units, the species, but the ecological units, the life forms. —G. Thorson (1957: p. 470) Though the technical difficulties are very great, they could probably be solved by anyone who really wanted to compare the furry growth of diatoms on a stone in a stream with the largerscale patches of woodland that have about the same sort of uniformity when viewed from an airplane. —G. E. Hutchinson (1965: p. 77) Is the modern marine biota composed of the same life habits as ancient ones? Which biotas are ecologically more diverse, in terms of both the number of life habits and the disparity (similarity) of these life habits? These are basic questions that ought to be answerable quantitatively by comparative paleoecologists. I will argue below that the answers to these and similar questions are impeded by a methodological limitation in our ability to compare communities (or other ecological entities) when they are separated by vast expanses of time and space and when they share few or no evolutionary homologies. Their solution hinges on the ability to compare quantitatively all kinds of entities directly on the basis of their ecological capabilities. Taxonomy has remained a typical yardstick for such comparisons. It has formed the dominant basis for comparing the structure of Paleozoic and Recent communities (Bretsky 1968; Ziegler et al. 1968; Walker and Laporte 1970; Levinton and Bambach 1975; West 1976; Miller 1988; Radenbaugh and McKinney 1998). Although all of these studies considered various ecological characters (e.g., trophic guilds, abundance), their primary impetus was the 274 PHILIP M. NOVACK-GOTTSHALL presence of taxonomically similar entities. The underlying assumption when using taxonomy in this way is that the ecological characters of taxonomic groups are conserved during evolution, such that taxonomy acts as shorthand for ecology. Although this may be generally true at low taxonomic levels, and occasionally high ones (Webb et al. 2002), there are many exceptions. For example, Fauchald and Jumars (1979) noted stark population-level differences within individual species of polychaetes, and Stanley (1968, 1972) and Miller (1990) noted widespread life habit convergence among bivalve orders. As a general rule, Peterson et al. (1999) demonstrated that conservatism is less likely above the familial level. Thus, although taxonomic comparisons may be suitable for documenting the ecological organization of taxonomically similar communities, such a basis is not useful when comparing taxonomically disparate communities. In short, taxonomy is an indirect, and potentially misleading, proxy for getting at ecological questions. Morphology has been another vehicle for ecological comparisons (Van Valkenburgh 1985, 1988, 1991, 1994; Foote 1996b; Wainwright and Reilly 1994; Van Valkenburgh and Molnar 2002; Lockwood 2004). The general premise of ecomorphology is that morphology can be used as a proxy for the ecological characters of organisms. Such correspondence has been well supported (e.g., Winemiller 1991; Wainwright 1994). However, there seems little potential in using these methods for large-scale comparisons spanning phyla and long time scales because of the lack of appropriate homologous characters. The most ambitious comparisons include Paleozoic and Recent arthropods (Briggs et al. 1992; Wills et al. 1994; Stockmeyer Lofgren et al. 2003) and animal skeletons (Thomas and Reif 1993; Thomas et al. 2000). There are few homologous (and even functionally comparable) morphological characters shared throughout benthic communities composed of green algae, foraminifera, corals, trilobites, bryozoans, brachiopods, and bivalves. It is essential to focus such comparisons on ecological characters directly, instead of on their underlying morphology or their consequences for taxonomy. It is important here to understand what I mean by the term ecological character. We can start with the understanding that each organism exhibits unique phenotypic features (sensu Bock and von Wahlert 1965) that affect environmental interactions. Collectively, these phenotypic features endow each organism with ecological capabilities or characters (faculties sensu Bock and von Wahlert 1965). For now, I will focus on those autecological characters related to feeding, use of space, mobility, dispersal, reproduction, and body size; taken together, these describe an organism’s basic life habit. Ecological diversity, regarded as the overall variety of life habits within some group, can be most easily assessed by richness, the number of unique life habits in this group. It can also be assessed by ecological disparity, a measure of how different each life habit is from others in this group (modified from Foote 1993a). I propose below a common framework for such characters and formal definitions for ecological richness and disparity. Focusing on such ecological characters directly has two benefits. First, it avoids the problems of homology associated with morphological comparisons. Because distinct phenotypes can perform identical functions in numerous ways (Bock and von Wahlert 1965; Alfaro et al. 2004, 2005; Wainwright et al. 2005; Marks and Lechowicz 2006), there exists in nature an innate tendency for ecological convergence when emergent capabilities are beneficial. In a sense, such higher-order capabilities are ‘‘screened off’’ (sensu Brandon 1984) from their underlying morphological and functional causes. These characters are accordingly more suitable for large-scale ecological comparisons. This may diminish, although not eliminate, the role of phylogenetic effects (Felsenstein 1985b; Harvey and Pagel 1991). Second, compared with analyses using the proxies of taxonomy or morphology alone, such a focus better aligns results with the theoretical understanding of ecological diversifications (Grant 1999; Schluter 2000; Coyne and Orr 2004). Such benefits motivated the development of the guild concept. Originally focused on comparisons among taxa sharing diet and foraging habits (Root 1967), it was later modified to 275 QUANTIFYING ECOLOGICAL DIVERSITY include other categories—microhabitat, locomotion, ecomorphology, timing of reproduction and daily activities, among others (Schoener 1974; Bambach 1983, 1985; Simberloff and Dayan 1991). Many studies have compared individual ecological characters over long time scales, including tiering (the stratification of infauna and epifauna; Thayer 1979, 1983; Ausich and Bottjer 1982; Bottjer and Ausich 1986; Droser and Bottjer 1989, 1993), insect feeding habits (Labandeira and Sepkoski 1993), energetic consumption (Vermeij 1999; Bambach 1993, 1999; Bambach et al. 2002), and body size (Smith et al. 2004), among others. Various individual characters related to escalation— chiefly carnivory, infaunality, and mobility— have been a recurrent focus (Vermeij 1977, 1987; Signor and Brett 1984; Kowalewski et al. 1998, 2006; Kosnik 2005; Madin et al. 2006; Aberhan et al. 2006). The most ambitious multivariate guild attempt was conducted by Bambach (1983, 1985) in a series of studies comparing the ecology of Sepkoski’s (1981) three evolutionary faunas. Using a three-dimensional framework defined by foraging habit, microhabitat, and mobility, he concluded that the timing of marine ecological diversification throughout the Phanerozoic was irregular and coincided with the diversification of successive evolutionary faunas, primarily resulting in increased utilization of previously vacant ecospace. These conclusions have withstood more recent analyses using broader ecological characters (Bambach et al. 2007; Bush et al. 2007). Bambach’s framework has been influential (Aberhan 1994; Bottjer et al. 1996; Droser et al. 1997; Radenbaugh and McKinney 1998), but different qualitative frameworks also exist. For example, the Evolution of Terrestrial Ecosystems consortium (Behrensmeyer et al. 1992, 2003; especially Wing 1988; Wing and DiMichele 1992; Damuth et al. 1992) used a comprehensive framework for comparing terrestrial communities. Retallack (2004) also presented a framework focused on general ecological strategies. Although such approaches are well suited to identifying synoptic ecological trends, they primarily are limited to making descriptive, qualitative comparisons or statistical comparisons of isolated ecological characters. A synthetic quantitative framework, while also allowing such analyses, is preferable for several reasons. First, it facilitates more robust documentation of overall changes in ecospace utilization (Bambach 1983, 1985, 1993). This can benefit our understanding of the previously mentioned univariate trends because their causes are likely intricately related to other ecological characters that this method captures simultaneously. Second, quantification makes it possible to determine the structural components of individuals occupying ecospace (Van Valen 1974). That is, it allows measurement of the central location, dispersion (disparity), and distribution of all individuals’ life habits in the multidimensional space defined by the ecospace framework. Of equal importance, it allows recognition of those ecological regions that are not occupied by individuals—either currently or in the past. Finally, quantification fosters the development of mechanistic null models that can test both the robustness of observed trends and distinguish among their possible causes (McShea 1994, 1998; Foote 1996a; Ciampaglio et al. 2001; Pie and Weitz 2005). The proposed framework marks the first framework suitable for such large-scale, quantitative comparisons. Such motivations drove the quantification of morphological disparity (Gould 1989, 1991; Briggs et al. 1992; Foote and Gould 1992; McShea 1993; Wills et al. 1994). Given the success of these approaches (Saunders and Swan 1984; Foote 1991a,b, 1992, 1993a,b, 1994, 1995, 1996a,b, 1999; Thomas and Reif 1993; Wagner 1995, 1997; Wills 1998, 2002; Lupia 1999; Smith and Lieberman 1999; Eble 2000; Thomas et al. 2000; Ciampaglio et al. 2001; Ciampaglio 2002; Harmon et al. 2003; Stockmeyer Lofgren et al. 2003; McClain et al. 2004; Villier and Korn 2004; Collar et al. 2005), quantification of ecological diversity seems to offer profound benefits. In this study, I propose a general method for quantifying ecological diversity that unites an extended framework of Bambach (1983, 1985) and the methodological advances of morphological disparity (see Foote 1991a; Wills 2002). The modified framework consists of 60 ecological character states that are universally applicable to extant or extinct organisms and 276 PHILIP M. NOVACK-GOTTSHALL that are logically independent of taxonomy; in this sense, the framework constitutes a theoretical ecospace. It allows quantification of ecological richness and disparity directly for any entity—individuals, lineages, or entire communities. The framework and the methods used in analyzing it are suitable for answering many questions in comparative paleoecology. Here it is used to compare the ecological diversity of Paleozoic (Cambrian through Devonian) and modern biotas from deep-subtidal, soft-substrate habitats in terms of ecological (life habit) richness, disparity, and overall distributions of life habit gradients in ordination-space. Paleozoic and Modern Data Sets The biotas used here represent assemblages from deep-subtidal, soft-substrate habitats. The Paleozoic biota comprises 449 samples compiled from 167 references, including nearly 80,000 individual fossils (an underestimate considering only one-quarter of samples have abundance data) and more than 3500 species ranging in age from Cambrian through Devonian (Novack-Gottshall 2004). The modern biota comprises 50 samples compiled from three references in the literature. Ten samples were selected at random from comparable habitats along the western North Atlantic— five samples from the Mid Atlantic Bight (Lynch et al. 1979) and five from the Beaufort Shelf (Day et al. 1971)—totaling more than 8000 individual organisms and 450 species from the Boreal Province on an outer continental shelf margin. Although these samples are from the same habitat as the Paleozoic samples, the temperate, oceanic shelf does not represent the same latitude as most Paleozoic samples. To account for this difference, 40 samples were also selected at random from appropriate habitats in the tropical, epeiric Gulf of Carpentaria (Australia) (Long et al. 1995), totaling more than 91,000 individuals and 400 species. The Ecospace Framework The life habits of the taxa in the biotas were operationalized by using the following standardized ecospace framework criteria. It is important to note that although the framework is well suited for comparing such marine biotas, it is equally well suited for characterizing the life habits of other ecological groups; the explanations that follow draw on examples from the full spectrum of life, both extinct and extant and representing most habitats. Characters in the Framework. The framework (Table 1; see also Appendix A online at http://dx.doi.org/10.1666/pbio06054.s1) includes 60 character states in 27 characters that describe the basic autecological capabilities of organisms. Characters include (1) resources, such as diet and microhabitat; (2) structures, behaviors, or other features related to the acquisition, maintenance, or defense of these resources, such as foraging, mobility, and substrate attachment; and (3) other important autecological characters, including body size, physiology, and reproduction. Depending on the scope of analysis, some researchers (especially macroecologists and paleobiologists) may be inclined to add geographic range, abundance, or other emergent (statistical, sensu Maurer 1999) group characters to this list (Peters 1983; Brown 1995; Gaston and Blackburn 1996; Maurer 1999). Adopting cladistic terminology, the term character refers to individual classes of ecological capabilities (faculties sensu Bock and von Wahlert 1965, whereas character state denotes the possible types of these capabilities (Swofford et al. 1996). The characters were chosen according to four criteria. First, the characters must be ecologically important for living organisms. Habitat and dietary characters are given greater emphasis—that is, there are more characters—because of their recognized importance (Schoener 1974). Second, the characters must be logically independent of one another; that is, they refer to different components of life habits. This is a requirement of all theoretical multidimensional morphospaces (McGhee 1999) and even cladistics (Swofford et al. 1996). In reality, correlations may exist and can be investigated a posteriori, but the assumption here is that all character combinations are possible—even if never realized because of constraints (Seilacher 1970). Third, the characters must be assignable to ancient taxa, including long-extinct species with no 277 QUANTIFYING ECOLOGICAL DIVERSITY TABLE 1. Twenty-seven characters (bold) and 60 states (numbered) in ecospace framework. Characters listed in parentheses are not easily determined for many fossil groups. Binary characters Reproduction Primary feeding microhabitat 1. Sexual 30. Above primary substrate 2. Asexual/Clonal 31. Within primary substrate (Development) Immediate feeding microhabitat 3. Direct development 32. Above immediate substrate 4. Indirect, non-feeding development 33. Within immediate substrate 5. Indirect, feeding development Diet (Brooding) 34. Autotroph 6. Brooding 35. Microbivore (Dispersal vector) 36. Carnivore 7. No vector 37. Herbivore 8. Fluid-dispersed 38. Fungivore 9. Organism-dispersed Physical condition of food 10. Self-dispersed 39. Incorporeal feeder Mobility 40. Solution feeder 11. Sedentary 41. Particle feeder 12. Passively mobile 42. Bulk feeder 13. Facultatively mobile Feeding habit 14. Intermittently mobile 43. Ambient feeder 15. Habitually mobile 44. Filter feeder Substrate/medium composition 45. Attachment feeder 16. Biotic 46. Mass feeder 17. Lithic 47. Raptorial feeder 18. Fluidic (Feeding selectivity) Substrate consistency 48. Non-selective 19. Hard 49. Selective 20. Soft 50. Secondarily selective 21. Insubstantial Ordered, multistate characters Substrate relationship 51. Skeletal body volume 22. Attached 52. Primary stratification 23. Free-living 53. Immediate stratification Primary microhabitat 54. Primary food stratification 24. Above primary substrate 55. Immediate food stratification 25. Within primary substrate 56. (Mobile velocity) Immediate microhabitat 57. (Spatial patterning) 26. Above immediate substrate 58. (Dispersal distance) 27. Within immediate substrate 59. (Relative metabolic rate) Support 60. (Life span) 28. Supported 29. Self-supported living relatives or morphological analogues. Reliance on taxonomic information has been minimized by focusing on general—and consequently often convergent—ecological capabilities of organisms instead of on particular, often taxon-specific adaptations. Fourth, the individual states for each character must be fully subdivided. For example, a fluidic substrate (Table 1) is a valid substrate state because it represents a logical absence of a substrate, used by organisms that do not inhabit lithic or biotic substrates; see Appendix A for further examples. The ecospace framework does not include synecological characters, except when an organism’s autecological characters necessarily imply some form of interaction. For example, carnivores are categorized only as meat eaters, and not with regard to their particular prey. In other words, character states referring to particular organisms—trilobite eaters, nectar eaters, and the like—were avoided because they limit comparisons to particular times when that dietary item was extant. This may limit the framework’s utility for some comparisons, but it is a prerequisite if comparisons are to be made across wide taxonomical, morphological, and ecological ranges. Modifications of this framework are possible depending on the objectives of the study. A comparison spanning the history of life on Earth might find the character states carnivore, her278 PHILIP M. NOVACK-GOTTSHALL bivore, and fungivore too restrictive; a replacement with chemoheterotroph might prove more useful. Unlike the Skeleton Space (Thomas and Reif 1993; Thomas et al. 2000), this list is provisional and not intended to be fully inclusive. Although it is applied below to marine biotas, it is intended to characterize universally the significant autecological capabilities of all organisms in any habitat. Additional characters can be devised when such information is available or when a study requires them. Reproductive strategies, seasonality, daily cyclicity, food size, and numerous biogeochemical and physiological characters are important ecologically (Schoener 1974; Pianka 2000), but this information is not available for most fossil species, and so it is not included in the present treatment. Some possible candidates are listed here (Table 1 and Appendix A, in parentheses) with the hope they will be included in future comparisons. Similarly, it may sometimes be necessary to limit the number of characters and states if relevant information is not available. In the examples that follow, for instance, only 44 character states are used because of the current limitations of using fossilized species. Coding of Character States. Unless noted, the term individuals in the following refers to individual ecological entities—individuals or species—whereas the term groups refers to more inclusive groups—communities and lineages. Most character states are binary, coded 0 for absent and 1 for present. Several characters—body size, microhabitat stratification and others—are coded as continuous, ordered, multistate characters by using integers (or fractions if de-weighting is preferred; Sneath and Sokal 1973; Van Valen 1974). Such multistate characters are used only when there is a clear ordination among their states. When a state cannot be confidently assigned or is unknown currently, it can be coded as unknown. Such codings, however, reflect only a lack of knowledge rather than nonrelevance; in principle, all states can be coded. This method of coding, in which multistate characters are the exception, might seem to warrant further explanation. In most cladistics or morphological disparity studies, the characters are typically homologous, with each individual displaying a single phenotype. In contrast, individuals are more variable ecologically, marked by behavioral flexibility, generalism, and convergence with unrelated individuals (Peterson et al. 1999; Losos et al. 2003). This can be notably true for sexually dimorphic species (e.g., Pietsch 1976, 2005). The coding scheme used here accommodates such variability by allowing single individuals to be coded with multiple states in the same character. For example, semi-infaunal individuals, such as trees and some mussels, can be coded as living simultaneously above and within the sediment. (Other common ecological and behavioral capabilities best described by multiple character states for the same individual—hermaphroditism, parthenogenesis, substrate and microhabitat generalism, omnivory, among others—are discussed further in Appendix A.) In cases where individuals typically utilize a primary resource, even when capable of using others, only the primary resource is coded; this is the same method used to classify guilds (Root 1967). A limitation of such flexibility in coding is that every individual must exhibit at least one state for every character. In other words, no individual—while alive—lacks some diet, some microhabitat, or some body size. This has analytical consequences, namely that not every combination in the framework is possible. Although individuals can undergo changes in their ecological characters throughout their life cycle—most notably due to metamorphosis or allometry—they are coded from the perspective of adult, sexually mature organisms, where ecological characters are usually most stable. Organisms with indeterminate growth are coded at the attainment of sexual maturity. Entire colonies are treated as individuals. Depending on the goals of a study, one could focus on each colony member individually, include individual genders or age or life stages separately, or code individuals within a population separately. Some characters, such as absolute body size and microhabitat stratification, are scale-independent and coded according to absolute criteria. However, because the ecospace framework has an autecological focus, most char279 QUANTIFYING ECOLOGICAL DIVERSITY TABLE 2. Utility of ecospace for describing benthic microhabitats. If using just three characters (primary microhabitat, immediate microhabitat, and substrate) with several character states (in italics), it is possible to describe twelve unique microhabitat combinations. Although existing terminology exists for most combinations, the three characters are more succinct and more broadly applicable for describing them. Additional combinations are possible; for example, it is possible to occupy multiple states simultaneously. The same classification can be used for other focal habitats; in this example, the habitat is the benthic one with the sediment-water interface as the primary substrate. See text for further discussion. This ecospace framework allows the discovery of combinations that are unoccupied in nature, such as the microhabitat that is within the benthic sediment but above the water; although this may seem an unlikely life habit, it might be possible to imagine an organism that floats or flies in gas-filled chambers within a burrow network. Primary microhabitat Immediate microhabitat Above S/W Above Within Within S/W