Closed regenerative life support systems for space travel: their development poses fundamental questions for ecological science

Estuon'ne Perspectives

ON THE EPISTEMOLOGY OF ECOSYSTEM ANALYSIS L. B. Slobodkln Department of Ecology and Evolution State University of New York Stony Brook, New York D. B. Botkin Enuironmentaf Studies University of California Santa Barbara, California B. Maguire, Jr. Department of Biology Un iversity of Texas Austin, Texas B. Moore, III Complex Systems Group O'Kane House University of New Hampshire Durham, New Hampshire and H. Morowitz Department of Molecular Biophysics and Biochemistry Yale (Jniuersity New Haven, Connecticut Abstract: It is impossible to construct a general theory or model of any particular ecosystem which will be useful for answering all possible questions about that system, although if we know enough about any ecosystem it is possible to construct such models once a specific question has been posed. This knowledge cannot be gained entirely from the system at issue, due to restrictions in time and resources, as well as to the fact that certain kinds of thorough ecological analysis may damage the system analyzed. Therefore, it is advisable to use relevant information from ecosystems other than the one of immediate interest. A partial list of species present in an ecosystem permits access to the information gained by naturalists working on other systems. We therefore justify the usual practice of making species lists because such a list Is the best (I.e. cheapest and most useful) preliminary step in answering questions about any ecosystem. While explicit measurements must also be made in the object ecosystem in order to usefully model it, it is likely that the number of such necessary measurements may be reduced and their usefulness enhanced by the background natural history information implicit

497

Copyright (S) 1980 by Academic Press, Inc. All rights of reproduction in any f o r m reserved. ISBN 0-12-404060-8

498

L. B. Stobodkin et al.

in a partial species list, To demonstrate that the information of natural history can be communicated In a relatively complete way, we provide a partial representation of an adaptive response surface for Hydra sp. in which much of the kind of information about these organisms that might be useful for model construction can be presented jn a relatively simple diagram.

The Problem for the Naturalist Ecology has progressed to the point that certain epistemological problems, i.e. problems in the theory of knowledge, of a non-trivial kind have emerged as real and pressing. We have a great deal of information about ecosystems. Obviously we can use more, but we can no longer use logical structures borrowed from Physics, Mathematics, or Computing Science as a framework for our knowledge without considering the unique properties of ecological systems. We will consider a somewhat simplified description of the relation between the naturalists, who are the source of detailed factual knowledge about ecosystems, and the mathematically oriented theoreticians, who are to a large degree responsible for the systemization which makes this knowledge generally accessible. To discover intimate facts about nature is extremely difficult, expensive, and time-consuming. Naturalists know that a fruitful lifetime of research can often be represented by a half-dozen facts about a few species, often with no assurance that they apply to other species. By contrast, the normal format of mathematical system construction in a physical or engineering context is first to provide an essentially complete definition of the system, in which it is made clear which portions of the world are, and are not, relevant to the problem at hand. An airplane or a tub of water may be the system defined. A state description suitable to the problem at hand is then constructed. The speed of an airplane, combined with relevant measures of wind velocity, load, and turbulence may be such a state descriptor as may the temperature and pressure surrounding the tub of water. A theory of airplane flight involves feeding the definition of the airplane, that is, its weight and metal structure and so on, and an appropriate state descriptor into a model which will describe how the state of the airplane will change with time or with alterations in ambient circumstances. When the theoretician concerned with an ecosystem requests from the naturalist a definition and a state descriptor he is often given a compilation of facts which is, by and large, grossly inadequate for most theoretical purposes. The theorist, perhaps with management waiting for an answer, can do no better than attempt to build a plausible approximation to what the definition and state descriptor might be and proceed with theorization and prediction in the usual way, as if his knowledge were adequate to the task. As we know, the results are often dubious. For various reasons, the problem cannot be solved meaningfully by head-on crash program methods. Consider that the problem of studying a

On the Epistemology of Ecosystem Analysis

499

small salt marsh cannot be alleviated by putting 200 naturalists to work on that marsh, since the repeated passage of their rubber boots would probably be a sufficiently great perturbation to destroy many interesting or valuable properties that the marsh might have possessed. We cannot even escape from our problem when critical definitional properties have been exceeded. Once an airplane crashes, its parts are a problem for someone other than the pilot but a drastically altered ecosystem is still a problem for ecologists. So far we claim to have demonstrated that a problem exists. The remainder of the paper will attempt to demonstrate one set of possible steps to the solution of the problem.

The Constraints on Ecological

Descriptors

We must distinguish between learning about an ecosystem and learn ing from an ecosystem. To learn from an ecosystem implies that we would use information gained from a particular system to build a model complete enough to permit answering a specific question about that system. Even if we replace rubber-booted naturalists with daintier information gathering systems, such sis electronic sensors, the construction of a complete model of an ecosystem, using only information from that system itself, is often difficult. First, data over time are required to model such systems, and regardless of the subtlety of data collection this time must be expended. Second, when attempts have been made to count the number of kinds of organisms in particular ecosystems (Elton 1966), numbers in the thousands are reached. A complete model of such a system would involve minimally a matrix of enormously high dimension, particularly when we consider that the number of kinds of between-species interactions may be quite large. It is useful for some purposes to consider the multiplicity of species and organisms in an ecosystem as statistical populations analogous with those of statistical mechanics {Kerner 1972). The relatively small number of kinds of particles dealt with in statistical mechanics and the possibility of assuming equivalance between samples taken from the system, i.e. the property of ergodicity, is a major contributor to the relative simplicity and generality of theories in physics and chemistry. By contrast, the parts of ecological systems differ from each other in so many interesting ways that we cannot assume the statistical equivalence of samples separate from each other in either space or time, except under very constrained circumstances, and we certainly cannot assume that a relatively small number of kinds of entities are interacting; therefore for most purposes an approach analogous to that of statistical mechanics is not available. If complete data could be collected without destroying the system under study, the construction of a complete model would be too difficult to handle by present computational methods, and, if it could be handled, the resulting model might be as difficult to interpret as nature itself. We conclude, in part, that if we are to understand an ecosystem we must bring in-

L. B. Stobodkin et al.

500

formation to that system from other sources. We omit discussion of the problem of verifying a model that was constructed from information taken from a particular ecosystem by using the same ecosystem in the process of verification. If we are to abandon the hope of making a complete description, what is the best possible descriptor available to us? From here on we will advance a positive program rather than make negative arguments. We can state some of the properties which a workable, realistic description of an actual ecosystem must have. It must be concise enough to be completely stated. It must contain a definition of the system so that we can tell different ecosystems apart. Jt must include information about the system, but it cannot constrain the questions that might be asked about the system. The last point is non-obvious. We do not know in advance what questions will be necessary to be asked about ecosystems, but we would like, if possible, to describe ecosystems in advance of the occurrence of these questions to facilitate our reply. We therefore would prefer a description which is theoretically neutral.

An Ecosystem

Descriptor

We now suggest how to construct such a descriptor, and then we will give an example of its use. We must first request a kind of tolerance from the reader. What we will present is neither difficult nor esoteric. It corresponds to the naturalists' sense of nature and uses the intuitions and attitudes of biology in a simple way. This simplicity is no reason to reject it, a priori, because it can be shown, despite our apparent reversion to 19th Century standards, that it is really the best system of description for actual problem solving. We suggest most strongly and seriously that the optimal definition available for any ecosystem is the list of species that has been found in that system. We will now defend this apparently trivial assertion. A species name, despite being subject to change as taxonomic revisions occur, is enormously rich in information in the sense that it provides access, through the scientific literature, to the work of naturalists studying organisms of that kind and similar kinds over the past three centuries. It also provides information about all of the properties these organisms share with the broader taxonomic groups to which the particular species belongs. This pool of information also includes statements about necessary environmental conditions and biological prerequisites of the organisms. In fact, there is no cheaper or more effective way of gaining flexible, useful information about an ecosystem than making a list of the species that can relatively easily be found there. We must immediately make it clear that we are not asserting that information about the individual component species in an ecosystem can replace observation of that system itself, since to understand ecosystems requires that we understand interactions of many kinds as well as understanding the interacting components. We are as-

On the Epistemology of Ecosystem

Analysis

501

serting that knowledge about the component species is extremely useful as a first step to the understanding of any ecosystem for any purpose. The total amount of information that may be available about the organisms in a species list varies but it may be so enormous and unwieldy as to make its usefulness suspect. This is not an unsurmountable problem since actual investigation of actual ecosystems are undertaken with a specific goal. The particular goal of any particular study defines the way the information associated with the species list is to be used and also dictates the subset of the total available information that need be considered. Consider what is learned from the first species named on the list. Of the approximately 2 million species on Earth we have found a particular one. We know roughly its temperature and food requirements, and we know whether or not it is terrestrial or marine or aquatic. We can therefore eliminate, as possible complicators of our consideration of this ecosystem, more than 1 million species, as having requirements incompatible with the known property of the organisms we have found. In almost all cases, finding that first organism had essentially no impact on the ecosystem itself. There are some ecosystems in which the acquisition of an incomplete species list may involve damage to the system of a fairly serious kind. These include attempts to make a list of the species living in coral heads or in soil. They also include situations in which rather rare species may be seriously damaged by the attempt to determine whether or not they are present. But in almost all cases, finding the first species of the species list has essentially no impact on the ecosystem itself. Now imagine collecting another specimen, avoiding taxonomic duplication. This second species will reduce the class of possible species yet to be found to the logical product of the species conceivably coexistent with both it and the first species, adding more to our list of impossible new finds. After 20 or 30 kinds of organisms have been found, the 21st or perhaps 31st new species may add very little new information about the characteristics of that ecosystem. At this point, the 2 million possible kinds of inhabitants that might have been present have been divided into four portions. The largest portion consists of organisms one can be sure are absent. Then there will be several thousand species that may or may not be expected to be found on further investigation. There is a third class of organisms that have not yet been found but we believe must be present, as their presence is a prerequisite for the occurrence of some organism we have actually seen. Finally, there is a small list of actually observed species. We also will have, from the data associated with past investigations of the observed species, a reasonably good idea of the physical and chemical properties of the ecosystem, probably a much better idea than we would have had if we had spent a corresponding amount of time and effort on chemical, geological, and meteorological analyses. The short species

502

L. B. Stobodkin et al.

list is therefore seen as a cheap, rapid, information-laden approach to defining an ecosystem. This procedure works for two basic reasons. First, each species is related to the world in a slightly different way, but all share the property of temporal continuity. Each species can be thought of as an environmental sensor, whose persistence in a particular place is a reliable indication of the condition and history of that place. To acquire knowledge of each species' properties would be prohibitively time-consuming, expensive and possibly environmentally perturbing, but we have free and quick access to information gained by past studies, generally in different ecosystems. Thus external information can be brought to the system in which we are interested. So far we claim a species list to be a uniquely valuable definition of any ecosystem for any purpose. The more complete the list the better, but even an incomplete list is very good. Note by contrast how little we know about the microstructure of an ecosystem if we know only its diversity index, biomass or energy flow. None of these would permit us to distinguish between a forest and a coral reef. If we conclude that a species list is a proper definition of an ecosystem, there remains the problem of how this definition may prove useful. For example, the relative abundance of species In a community also carries information. Compare a forest with one wild cherry tree and a thousand poplars with a forest that contains a thousand cherry trees and only one poplar! There are other descriptions which also each carry information of a different kind—flux diagrams of all kinds, biomass calculations, and so forth. Each of these is particularly suitable for some specific question about the ecosystem, but, as indicated above, we really don't know what question will be asked in advance and each of these descriptions may prove to be irrelevant to some set of questions. We claim that the information associated with a species list will be of value for answering almost any question that may be asked. In general, there is a distinction between defining a system and defining the state of that system. For most imaginable questions some state descriptor, in addition to the species list, is necessary, which will consist of a vector of measurements made in the ecosystem plus quantities, such as temperature range, which might be inferable from information about species in the observed species list. The general usefulness of a species list is of course dependent on how much we actually know about the individual species on the list so that natural history, i.e. basic information on the lives of the organisms concerned, is of general and permanent value. It also follows that a species list is less useful in areas where the preliminary natural history studies have not been undertaken. The general value of natural history and of the species list is not contingent on any attempt to model the ecosystem in advance of a specific question. It is our assertion that regardless of what question is asked in the future, this information will prove to be of value.

On the Epistemology of Ecosystem Analysis

^

r

503

What we have described is actually the way the best of ecosystem studies in ecological management programs have proceeded. All we have done is dissect out the methodological structure and recommend its general utilization. Further amplification of the rationale for this procedure and of the more formal properties of species lists and state descriptors in general can be found in Botkin et al. (1979) and Maguire et al. (1980). Once it is conceded that a species list plus a state vector are of enormously high information content, it is possible to consider the list itself to be an interesting object for theoretical study. Freckman et al. (1980) were concerned with how one might test the hypothesis that communities with larger numbers of species will differ in their response to perturbation from communities with smaller numbers of species. This relates to the concept of species packing and to other theories of diversity. In order to avoid preconceptions and selection of information, they considered it advisable to study communities which were not well known to anyone, so that no preconceptions were possible. Therefore they collected 14 samples of soil at different locations near Riverside, California and identified the genera of living Nematodes in each. The samples were divided into two classes; those with more than 12 genera (species-rich) and those with 7 or fewer (species-poor). A relative frequency count of the first approximately 100 live specimens was made for each sample. Each sample was treated as a separate community described by a species list and the differential count vector (a kind of state vector). The whole samples were then subjected to a nematacide until approximately 50% of the specimens were immobilized. The Nematodes were washed out of the nematacide and a new generic list and differential count was made for the motile survivors. Freckman et al. (1980) then compared the state of each of the communities before and after perturbation. They found no significant difference between species-rich and species-poor samples in the proportion of species eliminated by the nematacide. But, considering the differential species count as a vector in N space, where N was the initial number of genera, both the degree of rotation and Euclidian distance between the vectors was greater after perturbation if the species list was initially longer. This supports the theoretical assertion that species-rich Nematode communities are composed of more sensitive individuals than are species-poor communities. In this example, a taxonomic list and simple state vector could be used as a means to test a theoretical assertion on the assumption that it is the best descriptor available. It may prove to be the case that use of this descriptor may bring a new simplicity into some kinds of ecological theory.

Complete Description In Natural History One of the assumptions in the above discussion was that one can actually start with a species' name and use it as a starting point for a search

504

L. B. Stobodkin et al.

of the scientific literature to discover what naturalists have learned about that species. Running counter to this assumption, however, is what might be called the naturalists' mystique, namely, that there is an esoteric quality to the knowledge of naturalists. If what is known by a naturalist cannot be stated in a reasonably complete and clear form, it is not very valuable by normal scientific standards. There exists a very real problem of making sure that the assertions of naturalists are publicly verifiable. We believe for two reasons that lengthy species-by-species descriptions are not necessary. The obvious reason is that, by and large, the classical system of classification works reasonably well, so that while certain details of muskrat biology may be the esoteric knowledge of only an experienced muskrat expert, many of the properties associated with rodents, mammals, chordates and eucaryotes in general are well known. We will not belabor this point further. The second, more speculative point we will illustrate with an example. Slobodkin (1980) attempted to summarize studies of various aspects of the biology of hydra in as complete and explicit a way as possible. He started with several facts, valid for some hydra, although they have not been demonstrated for all species (support for these assertions is presented in Hecker and Slobodkin (1977), Otto and Campbell (1977) and Slobodkin (1980) and in references cited therein): 1. There is an asymptotically increasing relation between food supply and both growth rate and budding rate at any particular temperature, kind of food, and water chemistry; 2. Between species, there is an inverse relation between growth rate and budding rate. This is intimately related to the apical control mechanism. That is, if there is strong dominance by the hypostome, budding is restricted and the food that would have supported the budding process can only be used for growth; 3. Increasing temperature lowers body size and increases budding rate for several species. It may, therefore, rotate the response surface towards small body size and rapid budding. Alterations in water chemistry may also distort the surface in a reasonably simple manner. From this information a partially hypothetical adaptive response surface for hydra can be constructed. This is presented in Figs. 1 and 2. The three axes are food level and the body size and budding rate attained by individual animals all at constant temperature and kept at constant food until an approximate steady state in their budding rate has been reached. The surface flattens against the food level-body size plane at low food levels, since hydra can be maintained at food levels which permit survival but not budding. At food levels below A, hydra die. As food level increases, the smaller species of hydra begin to bud before the larger ones. At sufficiently high food levels, the growth and reproductive rates approach an asymptote. The meridional lines (As,, etc.) on Fig. 1 represent the assumed trace of particular genotypes on this surface. That these lines do

"

*

On the Epistemology of Ecosystem

Analysis

505

A

Figure 1. The adaptive response surface for hydra. On the basis of preliminary information it seems likely that the interaction between food level, budding rate and body size in hydra can be represented by this surface. Meridional lines (s^ on this three dimensional surface represent the response of different genotypes of hydra. For further description see text

not cross is an explicit hypothesis. The surface shown in Fig. 2 is identical with that shown in Fig. 1 except that the meridional lines have been omitted for clarity. Closed regions have been mapped onto Fig. 2 indicating particular properties of hydra. For example, green hydra all bud rapidly but are relatively small. Floating occurs in relatively large species only, and only when they are hungry. Sexuality occurs primarily in intermediate sized hungry animals. Large well fed animals are predacious on small green animals. Slobodkin (1980) has suggested that essentially all of the ecologically interesting properties of hydra may prove to be mappable onto closed discrete regions of Fig. 2. If this is so, then at least in this case the possibility of a simple presentation of the complete natural history of a group of animals will have been demonstrated. The surfaces drawn in Figs. 1 and 2 are adaptive response surfaces which arise purely from observation of hydra. They are not a consequence of any more general theory. That is, no optimization theory or theory of population genetics or niche theory or adaptive landscape theory predicts how to draw the surface, although there is reason to suspect that a surface of this type is possible in general. While it is possible to rationalize the par-

506

L. B. Stobodkin et al.

Figure 2. Mapping of ecologically significant properties onto the response surface. For further description see text.

adaptive

ticular responses, the rationalizations are independent of the description. That is, the transmission of information about hydra is free of optimality assumptions or models, but can be used to infer optimizations or models. We have just provided the information to permit the reader to "think like a hydra" to some extent.

Conclusions We have suggested that it is both feasible and interesting to attempt to describe nature in the most information laden way possible on both the levels of the ecosystem and of natural history. The properties of the natural world may then dictate to us not only the facts of ecology, but also the form of ecological theory. We have given a few brief examples of particular descriptions. We believe that the procedure of attempting to produce rich descriptions is not only useful but that it generates important theoretical questions, different from those generated by attempting to fit ecology into the epistomelogical nexus generated by the physics and engineering of inanimate systems. This is not vitalism but is a refusal to relegate our intellectual problems to strangers.

On the Epistemology

of Ecosystem

Analysis

507

Acknowledgments Interaction between the co-authors was made possible by the support of NASA Space Biology Program. The studies of Hydra sp. from the laboratory of L. B. Slobodkin were supported by the National Science Foundation Ecology program. Contribution No. 297, Department of Ecology and Evolution, State University of New York at Stony Brook.

References Cited Botkin, D., B. Maguire, B. Moore, H. Morowitz and L. B. Slobodkin. 1979. Closed regenerative life support systems for space travel: Their development poses fundamental questions for ecological science. (COSPAR) Life Sciences and Space Research. XV1I:3-12. Elton, C. 1966. The Pattern of Animal Communities. Methuen, London, 432 pp. Freckman, D. W., L. B. Slobodkin and C. Taylor. 1980. Pesticide use and the stability of species-rich and species-poor communities of nematodes. Proc. VII Int. Soils Zoology Colloquium (in press). Hecker, B. and L. B. Slobodkin. 1977. Responses of Hydra oii'gactis to temperature and feeding rate, pp. 175-183. In-. G. O. Mackie (ed.) Coelenterate Ecology and Behavior. Plenum Press, New York. Kerner, E. 1972. Gibbs Ensemble, Biological Ensemble. Gordon Breach, New York. Maguire, B., D. Botkin, B. Moore, H. Morowitz and L, B. Slobodkin. 1980. A new paradigm for the examination of {closed) ecosystems, /n; J, Giesy (ed.), Symposium on Microcosms in Ecological Research, Technical Information Center, U.S. Dept. of Energy (in press). Otto, J. J. and R. C. Campbell. 1977. Tissue economics of Hydra: Regulation of cell cycle, animal size and development by controlled feeding rates. J. Cell. Sci. 28:117-132. Slobodkin, L. B. 1980. Problems in ecological description, I, The adaptive response surface of hydra. Mem. Inst. Ital. Idrobiol. (Suppl.) 37:77 95.