ISSN 1320-0682
|
|
|
|
ISSN 1320-0682 |
|
| Source: | http://www.complexity.org.au/ci/vol06/green/green.html | Received: | 01/07/1998 | ||
| Vol 6: | Copyright 1998 | Accepted for publication: | 15/10/1998 |
David G. Green and Nicholas I. Klomp
School of Environmental and Information Science
Charles Sturt University
Email: david.green@infotech.monash.edu.au
WWW:
http://www.csse.monash.edu.au/~dgreen/
The application of information technology to environmental issues is changing both theory and practice. The idea of ``natural computation'' provides new ways to understand environmental complexity across the entire range of scales, from individual phenotype to biogeography. Understanding the ways in which local interactions affect the global composition and dynamics of whole communities is crucial to the viability of strategies to manage ecosystems, especially in landscapes altered by human activity. Also environmental planning and management are increasingly dependent on accurate, up-to-date information that sets local decisions within a global context. The Internet makes it possible to combine environmental data from many different sources, raising the prospect of creating a global information warehouse that is distributed amongst many contributing sites.
Humankind is in the midst of a crisis. For thousands of years people have exploited the environment as though it were an infinite resource -- unchanging, predictable and inexhaustible. However the impacts of human activity are now felt everywhere. Conserving the world's flora and fauna is one of the great challenges of our time. Loss of biodiversity, ecosystem degradation and pollution are just some of the environmental problems on planet Earth. With human population and industrialisation still increasing rapidly, it is becoming vital to place a check on these problems within the next few decades.
In the face of this rapidly changing situation, traditional ideas and approaches to environmental management are no longer enough. To manage (say) a national park adequately requires knowing much more than simply what is happening within the park. It demands that local issues be set in the context of the surrounding region, as well as national and international developments, global change, socioeconomic influences, and a host of other issues as well.
As planners and managers learn to cope with this new scenario, we are witnessing the development of a new paradigm that integrates traditional field ecology with modern technology. It is a paradigm that links scientific research to environmental planning and management. It links diverse and potentially massive sources of information, from field ecology to satellite imagery. Such a new approach can be invoked to address a host of practical problems, from land use planning to global warming.
In this brief account we try to achieve three goals. The first is to explain briefly the nature of complexity in the environment. Secondly, we argue that a new paradigm -- environmental informatics -- is emerging out of responses to the growing need to cope with this complexity. Finally we sketch out some of the ``grand challenges'', both in research and in practice, that environmental informatics needs to address in the new millennium.
Even the simplest ecosystems are highly complex. Complexity in the environment is present for many reasons, but most many sources of complexity can be grouped into the categories described below.
Many influences of the world's environment come from sources outside the Earth's biosphere. The sun, the moon, meteors and geomagnetism all influence life on Earth. But even if considering only our biosphere, the sheer scale involved in global environmental management is immense. The planet's surface area totals over 509,000,000 square kilometres. Simply monitoring one factor (say, surface temperature) across such vast tracts is a huge task [27]. Thorough monitoring of all environmental factors or even rudimentary research of the entire surface of the Earth is currently impossible. Whilst modern technology can help (e.g. remote sensing), it generates huge volumes of data that must somehow be stored, collated and interpreted [19].
Many environmental processes occur over geological or evolutionary time. Even successional or micro-evolutional processes usually take place over time periods much longer than a human life (or the length of a typical ecological research project of 1-3 years!). This has led to many inaccurate, ``time-blinkered'' assumptions in ecology, such as stable community structures, climax states and other conclusions about balance and equilibria (see later).
Taxonomists have described about 1.5 million species [31]. The total number of species is not known, but is estimated to be somewhere between 10 million and 100 million. At the current pace it would take at least another 300 years of taxonomic research simply to document them all.
However it is not sheer numbers of species that make the living
world complex, but rather the enormous variety of ways in which
they combine and interact. For instance, suppose that 100 species
inhabit a region; then there are 4,950 possible pairs of interacting
species. However, when we look at possible combinations, the
possibilities blow out to astronomical proportions. There are
over 
ways in which we can select communities of
10 species at a time. For communities of 50 species at a time,
this number rises to over
combinations. This complexity
increases further by orders of magnitude when the interactions of biotic
and abiotic factors within an ecosystem are considered.
An important example of complexity, especially in landscapes, is the phase change between connected and fragmented population distributions [7]. For instance, if we remove small patches forest from a landscape then the forest as a whole retains its integrity. However if clearing continues (at random), then instead of small patches breaking off, the entire system remains connected until a critical point, whereupon it breaks down into many isolated fragments [8]. Such criticality or abrupt phase changes have now been documented in many natural systems [2], from pest and disease epidemics [10] to fire behaviour in forests [7].
One of the most important results to spring from ecological transect studies is that environmental factors alone do not fully explain the spatial distributions of organisms. For instance, competition between species often truncates distributions along an environmental gradient [25]. These results imply that ecosystems are not controlled in simple linear, fashion by external (i.e. abiotic) factors, but by interaction of biotic and abiotic factors within a system (eg. [2]).
Networks of interactions between species are a major source of complexity in ecosystems. Interactions between pairs of species can take many forms, such as predation and competition. Feedback loops are especially common in multi-species systems. In populations with seasonal reproduction, delays arising from feedback tend to produce cyclic behaviour. They can also lead to non-linear and chaotic dynamics [23].
Complexity in an ecosystem (as measured by species richness) does not necessarily imply stability [21, 22, 23]. One reason for this is that any random collection of interacting species is likely to contain at least one positive feedback loop, which destabilises the system and leads to local extinction of one or more species [28].
Human influences on ecosystems tend to be disturbances that disrupt any semblance of equilibrium. Two of the most far-reaching of these disturbances have been land clearing and the introduction of exotic species. In many cases the effects are unintended side effects. Examples include wildfires, spread of diseases, pollution, salinisation, and desertification, to name just a few. In every case the disturbances force ecosystems away from equilibrium and can lead to local extinctions or other abrupt changes (e.g. [7, 8]).
In addition to making ecosystems more complicated (or more difficult to manage), humans influence the management of ecosystems by directing goals and agendas in ways that require environmental management decisions to be based on much more than ecological knowledge. This has led to the recent, rapid increase in the use of decision support models by land managers [19].
Although still in its infancy, complexity theory holds some important lessons for environmental science and management. Only some of these have been widely recognised so far. Taken together they highlight the need for new ways of doing research and management in ecology. Here I briefly summarise some of these lessons.
Although reductionism has served science well, we must recognise that it fails badly in trying to make sense of environmental processes. Above we highlighted a few examples of the ways in which interactions between different populations can have unexpected effects. It has long been common practice in ecology to study individual populations separately (``autecology'') , without reference to the ways they interact with other populations. Other reductionist practices include breaking down community level dynamics into studies of physiology and other responses at the level of individuals.
A good example is the way in which dispersal (interactions between sites in a landscape) can affect the dynamics of whole ecosystems [8]. Rare species tend to form clumped distributions which help them to persist in the face of superior competitors. Simulation studies suggest that this process provides one mechanism which maintains high diversity in tropical rainforests. Field studies confirm that rainforest contain just a few common, widespread species, and many rare species [14]. All of these rare species have clumped distributions. By not fully understanding such dynamics, conservation, management and research can be rendered ill-conceived or ineffective.
Sensitivity to initial conditions is a well-known phenomenon in non-linear systems, and one of the hallmarks of chaos. It is especially common in ecology where so many interactions are non-linear [23]. As an example, consider what happens if the connectivity of a landscape is near the critical region mentioned earlier. Under such conditions the size and composition of connected patches becomes extremely variable so the outcome of processes that involve spread through a connected patch, such as fire, epidemics, and invasions, become inherently unpredictable [8, 10]. Likewise the addition of a single exotic species to an ecosystem alters the web of interactions between species, perhaps creating a potentially devastating positive feedback loop where none existed before [28].
The need to cope with unpredictability highlights the importance of tools such as simulation models. Simulation allows us to carry out virtual experiments. In environmental management such experiments are often impossible to carry out in practice, either because they would take too long (e.g. forest succession) or because they would be too damaging (e.g. burning down an entire forest). Although exact prediction may be impossible simulation makes it possible to examine ways of dealing with many potential scenarios.
The idea that nature is in equilibrium - a cornerstone of much thinking within the environmental movement - arises from several sources. Perhaps the most important is the exceedingly long time scale of many processes in forest ecosystems, as mentioned earlier. Individual trees often live for many hundreds of years and simple succession - the replacement of one community by another - can take literally thousands of years to complete [6]. The fact that forests change so slowly gives the false impression that they are in equilibrium.
The equilibrium assumption underlies many ideas in theoretical ecology. For instance, Macarthur and Wilson's theory of island biogeography suggested that for any island there is an equilibrium number of species that it can sustain [20]. However growing understanding of the large scale and long-term dynamics of ecosystems make equilibrium assumptions increasingly untenable. For example Clements' theory assumed that succession leads to an equilibrium climax state [3]. This theory dominated plant ecology for most of the twentieth century, until evidence accumulated for other kinds of dynamics, such as chronically disturbed ecosystems [24] and long-term instabilities in vegetation history [5, 6].
Perhaps more importantly the urgent need to address environmental management, especially in disturbed ecosystems, is forcing ecologists to search for non-equilibrium models.
Combining different datasets together often leads to unexpected discoveries. That is serendipity occurs. The probability of serendipity increases exponentially with the number of different datasets available. So large repositories of data are almost certain to be rich sources of new insights about environmental processes [9].
A new paradigm requires a new way of looking at the world. The increasing use of computers has stimulated a view in which the natural world is seen as a form of computation. The analogies are compelling. DNA is the code for life's ``program''. Organisms are akin to robots or agents, and animal communication is a form of information processing.
The links between biology (including ecology) and computing have been growing ever closer. Techniques such as genetic algorithms, cellular automata and neural networks clearly borrow on biological ideas. We have argued [10] that many algorithms can be improved by mimicking living systems more closely [18].
One of the major challenges for ecology is to bridge present gaps in our understanding in the spectrum of genotype, phenotype, population and community. Perhaps the least well understood is the link between genotype and phenotype, and thence to environmental processes. The obvious analogy for scholars of computing and complexity is that to understand how a computer program works it is not enough to understand what each line of code means. You also need to know how those lines of code are organised.
At present very little is known about the relationship between genetic composition and growth processes. Kauffmann [17] modelled genetic control over growth as a switching circuit in which genes are ON-OFF switches that not only code for certain proteins but also affect other genes. However there has been very little other work of this kind.
L-system models [26] are now so sophisticated that they can faithfully reproduce the potential growth form of many plants. Virtual plants are now being used to carry out virtual experiments and could help to bridge the gap between laboratory experiments and field observations. A crucial step is to understand the link between growth form and taxonomic relationships. That is, how do genetic variations impact on the models?
One of the most relevant and important developments associated with natural computation is a new research field called artificial life (`Alife' for short). This is the study of life-like properties in computational systems.
One of the key ideas in Alife is that of an agent. An agent is a discrete entity that has certain computational capabilities, and can also interact both with its surroundings and with other agents. An important area of Alife research, and of advanced computing generally, is to study the properties and behaviour of multi-agent systems. This research is beginning to grow into a significant body of theory about systems of this kind.
For instance, in one early study, Hogeweg and Hesper (1983) showed that the observed social organisation of bumblebees arises as a natural consequence of the interaction between simple properties of bumblebee behaviour and their environment. For example, one rule they invoke is the TODO principle [12, 13]. Bumblebees have no intended plan of action, they simply do whatever there is to do at any given time and place. Similar interactions lead to order in many other animal communities, such as ant colonies and flock formation by birds.
For most of the Twentieth Century, conservation could be equated with national parks. However the rapidly growing scale of environmental alteration and increasing public awareness of environmental issues have highlighted the need for off-reserve conservation and environment management [4]. The range of off-reserve issues is now very broad. Some examples include: environmental impact assessment, state of the environment reporting, environmental monitoring, conservation of rare and endangered species, natural heritage planning, species relocation programs, land use planning, and environmental degradation.
Out of all the above activity has emerged an awareness that local decisions and priorities need to be set in a wider, and ultimately global context [30]. For instance to decide whether to log a patch of rainforest, you have to know how much other rainforest there is, what species will be put at risk, what the global costs and benefits are, etc. Conversely, every local area contributes data and experience that can be applied to other areas and can feed into setting global priorities and policies.
The new paradigm that is emerging treats environmental management as a host of activities all of which reinforce each other. Each area of activity is both enhanced and constrained by the global picture. The key to the success of the new approach is this two-way communication. Setting matters in context means having access to relevant and reliable information. During the 1990s governments have been very active in setting up regional, national, and international environmental information systems (e.g. [1, 9, 30]).
The growth of the Internet has played an integral part in this emerging paradigm. Up until recently most research was carried out as a series of isolated studies. However, by sharing data over the Internet, the results of previous studies can enrich subsequent research. The best examples are in genomic research, where the development of large, on-line databases means not only that new sequences can be interpreted by comparing them with whole families of existing data, but also that entirely new kinds of studies are possible in which researchers mine the databases for unsuspected patterns and relationships. The challenge for ecology is to mobilise data from previous studies in similar fashion.
The essential advantage of the Internet (especially the World Wide Web ) is its ability to combine information from many different sources in seamless fashion [9]. This has created an unprecedented opportunity for data sharing and cooperation on scales that were formerly deemed impossible. It also brings sharply into focus the need for coordination. The explosive growth of the Internet has led to massive confusion. Many organisations are duplicating facilities in inconsistent ways. There is an urgent need to develop for agreed protocols and standards regarding, data recording, quality assurance , custodianship, copyright , legal liability and indexing [9].
One of the most urgent needs is to develop a consistent framework for discussing environmental issues. One of the most basic problems is that we do not even have a comprehensive list of the world's species. Not only that, the taxonomic nomenclature has been confused and inconsistent. It is not surprising then that some of the first initiatives in on-line environmental information have focussed on putting together consistent reference lists of the world's species. For instance since 1993 the International Organization for Plant Information (IOPI) has been developing a checklist of the world's plant species [15]. This is now contributing to recent major initiatives in this area, including the Species 2000 Project [16] and the Global Biodiversity Information Facility (GBIF), which are international projects of the OECD's Megascience Forum [11]. The aim is to establish ``... a common access system, Internet-based, for accessing the world's known species through some 180 global species databases ...''
A major challenge is to flesh out and complement the data that is now available with facilities that allow people to use it effectively. Along with data warehouses, we also need information systems to interpret and apply the information. For instance, foresters, faced with the need to demonstrate the environmental impact of logging operations, have developed simulation tools such as the visualisation program SmartForest. This program [29] integrates simulation models with geographic information to create views of future landscapes under selected scenarios.
Learning to conserve the world's living resources is one of the great challenges of our time. In a very real sense the future of humanity depends on finding a solution. It is not an easy problem to solve.
As we have seen here, achieving these goals will demand a much better understanding of environmental complexity than we have at present. Thus there is a need for greater dialogue between ecology and complexity studies. At present the extent of this dialogue is still small. With a few notable exceptions, most ecologists are largely unaware that the field of complexity even exists, and many researchers in (say) Alife are computer scientists who are unaware of the major issues and questions driving ecological research.
We can no longer pretend to manage nature in isolation from human activity. Human activity has expanded to affect virtually every ecosystem, everywhere. We have to learn to manage ecosystems that are not only out of equilibrium but also chronically disturbed and largely unpredictable. We can no longer confine conservation to ``isolated'', ``natural'' parks. Conservation needs to incorporated into the ways we deal with all living systems in all environments.
Global conservation demands a much greater level of coordination than at present. This coordination includes two-way communication between the activities of different conservation agencies and groups. It also implies much greater planning because almost every socioeconomic activity potentially impinges on conservation. To achieve both of these ends, greater dialogue between ecologists and computer scientists is needed urgently.
 |