Introduction

There is absolutely no question that the biosphere is being altered in a substantial manner by the activities of humans. Changes in the configuration of land surfaces, massive rearrangements of the earth's biota, and an atmosphere that is increasing in concentrations of a number of radiatively active gases are but a few signs indicating the large magnitude of these alterations. Predicting the ecosystem consequences of these changes is a formidable challenge. How will, for example, the increasing concentrations of CO₂ influence ecosystem functioning and distribution? What kinds of tools do we currently have to even approach an answer to such a question?

    As a first approach we can look to the historical record. There have been changes in CO₂ concentrations of the atmosphere in the past and accompanying biotic shifts. However, the historical information available is at a fairly coarse resolution, does not provide direct information on system functioning, and does not give clues as to what to expect when atmospheres are not only enriched in CO₂ but also in ozone, sulfur, and nitrogen gases, as we are seeing today. Also the historical rates of change are slower than recent and projected changes in the environment.

   A second approach available to us is the use of simulation models that are built for the most part on information on the responses and reactions of pieces of ecosystems-e.g photosynthesis of leaves in response to CO₂ increase, increase in decomposition due to increased temperature, and so forth. These models are enormously revealing in pointing to the sensitivity of various component processes to environmental changes, as well as indicating where further information is needed in order to understand system interactions. We have learned, however, that there is an enormous 'buffering' capacity as well as compensations within components of systems such that simplistic predictions of responses to change derived from these models may be misleading even in relatively short time periods.

  There is a third approach for deriving an understanding of the consequences of change--experimental manipulation of whole ecosystems. Such experiments can tell us how whole systems, as well as their components, will respond to change both in the short and long term, as well as providing direct tests of the models discussed above.

  The utilization of ecosystem experimentation has been immensely rewarding, yet has received limited application, in part because of difficulty of replication, and the large expense and long duration of these experiments. However, as the need for knowledge of ecosystem response to change becomes increasingly more urgent we must begin to apply this direct method more widely to complement historical and modelling approaches.

  The objective of this volume is to explore the potential of ecosystem experimentation as a tool to understanding and predicting more precisely the consequences of our changing biosphere. We take a broad view of the problem by first examining what we have learned from 'natural' experiments as well as large-scale inadvertent ecosystem perturbations induced by human action. Then the results of explicit ecosystem manipulations are reviewed and finally suggestions are made of the kinds of large-scale experiments that might be useful for the future, as well as an analysis of the available tools to do so.

  What can we learn from 'natural' experiments? Trillmich's review of the impact of the 1982-83 El Niño on the Galapagos Islands is revealing. Even though there have been no ecosystem studies as such on these islands there have been intensive population studies of representatives of the various trophic levels, in some cases extending back many years prior to the major climatic perturbation that brought 10 times the normal rainfall. Trillmich is able to piece together the interactions of the impact of the perturbation on and among representatives of the various trophic levels. He concludes, though, by noting that without population measurements extending some years prior to the perturbation it would be difficult to interpret the events immediately following the event. Further, the impact will have long-lasting effects that will require continuing observations to fully document and evaluate.

  What can we learn from the 'inadvertent' experiments that have resulted principally from the activities of humans on natural systems. A series of papers in this volume address this issue: Medina for the tropical forests, Tamm for temperate forests, and Graetz for desert systems. All take a different approach to the problem. Medina reviews the information on the impacts of deforestation on nutrient, carbon and water cycles, and shows how these manipulations are having profound influences in the short and long term. These analyses provide insights into the mechanisms involved; however, they are not definitive since any particular study generally lacks adequate long-term pre-disturbance measurements as well as controls, and they most often do not view comprehensive system responses. Medina outlines specific system manipulations that would, however, lead to a capacity to predict the consequences of the kinds of modifications that are, and will be affecting tropical forest ecosystems.

    Tamm quite naturally takes a more historical approach to the analysis of the human impacts of disturbance on temperate forests, since such perturbations have been occurring for millennia. He shows how various tools can be utilized to evaluate the effects of these ancient disturbances. He points out, however, that it is often difficult to detect the signal of human-induced disturbance from that of climate change. He calls for a better-developed theory of ecosystem function in order to utilize more fully historical techniques.

   Arid systems, because of their relative simplicity and close coupling to rainfall events, offer the possibility of clearly revealing not just the effects of humans on these systems, but the interaction of human systems with unusual climatic occurrences. Graetz is thus able to unravel the interactive result of a severe drought in the Sahel on a nomadic society and the arid ecosystem of which it is a component. It would be difficult to design explicit studies of ecosystems that would include the interactive influences of societies, although such information is highly desirable.

  Schulze, Ulrich, and Herrmann take on the difficult task of evaluating the ecosystem response of exposure to chronic pollutants. Schulze and Ulrich show how the effects of acid rain have been progressing for a long time in conjunction with other influences on system functioning. The correlative findings provide the framework for future explicit experiments but in themselves are not totally conclusive. One thing is clear from their analysis--a single atmospheric pollutant can have effects that ramify through a whole ecosystem, and because of buffering and feedbacks, cause and effect relation-ships may be very difficult to unravel, without experimentally probing various parts of the network.

   Herrmann, examining varied pollutant inputs and transfers within ecosystems, concludes that the unfortunately too common uncontrolled experiments give us but crude information on rates of transfer processes. He proposes a combination field, laboratory, and modelling effort to more fully utilize the information that we have in developing an understanding of the dynamics of the movement of pollutants within ecosystems.

  Those contributors to this volume that assessed the inadvertent, uncontrolled experiments that humans have been performing with increasing intensity and frequency all call for the development of a better ecosystem conceptual base in order to evaluate and understand the complex responses that are noted. They think that, in addition to the development of a more robust ecosystem theory and increased model development, small-scale experiments are essential for understanding the functioning and responses of particular parts of the system.

  What about explicit whole-system experiments? Schindler makes a powerful case for the utility of such experiments. He demonstrates that issues that were unable to be resolved by the traditional observational and correlative approaches to the study of natural systems were resolved with relatively simple experimental designs. Yet each experiment seems to give some surprising results, underscoring again the poor state of our knowledge of the comparative functioning of ecosystems. Schindler echos the plea of Trillmich for long-term observations in order to understand fully the responses of systems to perturbation, as well as giving many guidelines for whole-system experiments that he has developed over two decades of experience. He urges that the manipulations be designed to yield widely generalizable results, that they be accompanied by intensive comparisons with smaller-scale experimental or observational techniques that have the possibility of being used more widely. Finally, Schindler gives some very explicit experimental approaches that might be utilized to predict the consequences of global change on northern aquatic systems.

   Moving from the comparative simplicity of lakes that Schindler considers to the more complex interface of land and water systems, Bayley and Schindler show how fire can be utilized as a tool to test some specific hypotheses about the regulation of nutrient and water transfer in perturbed systems. Pre-disturbance studies, as well as long-term observations subsequent to fire, made it possible to answer the questions that were posed even in these relatively complicated systems.

   Both Wright and Attiwill discuss the utilization of watersheds for ecosystem experiments. Attiwill shows how watershed manipulations have been applied first to understanding system water balance, and subsequently to nutrient balance. He, as well as Wright, discuss the problem of suitable references for such manipulations in which there is often but a single manipulated watershed and a reference untreated forested watershed. This brings up an issue which is addressed more fully in subsequent chapters, and one which is the greatest problem to overcome in the full application of ecosystem experimentation: how to get adequate replication in manipulations that are often very large and generally quite expensive, if for no other reason than the length of time that the measurement series must be continued after manipulation in order to get meaningful results. The report of the aquatic group led by Kitchell addresses this issue, and contends that new statistical tools are available that can, as the group state, 'overcome the constraints of statistical tradition'. This is an issue that needs intense consideration since, from all other indications, the experimental approach to ecosystem science is crucial to bringing our understanding of ecosystem functioning to a new level. Wright gives some practical advice on experimental design of watershed manipulations.

   Recently ecosystem-level experiments have begun to examine the influence of atmospheric changes on system function. Wright shows the results of whole watershed acidification experiments, and makes the important observation that 'controlled experiments may offer the only realistic approach to quantifying the effects of future changes that have as yet no modern analog'.

   For aquatic ecosystem experimentation the lake has been the most utilized experimental unit, and for terrestrial systems the watershed. What about wetlands? Ewel indicates that basins have been useful but that plots can also be utilized in combination with modeling approaches realizing the limitations of interpretations to larger scales. A further useful approach is the employment of constructed hydrologically isolated 'cells' within the wetlands, analogous to the 'mesocosms' discussed for lakes by Schindler. Each approach has its merits and limitations, and their application depends on the question being posed and the resources available.

  The issue of the size of the experimental unit is discussed by a number of the contributors, including Schindler, Wright, and Ewel. Schindler discusses at some length the different sorts of interpretations that can be made depending on the experimental unit size. The problem of scaling has received increasing attention during the past few years as ecologists are considering how to make their findings available to scientists who generally work on larger scales than lakes or watersheds, such as atmospheric scientists. There is no doubt that the scaling issue will receive increasing attention in the years ahead, as more and more disciplines interact in the study of the consequences of global change. Levin contributes to this discussion by showing how mathematical techniques can deal with interpreting processes across large scale dimensions.

   The various contributions to this volume thus illustrate the enormous difficulties of interpreting ecosystem function, and response to change, by observational and correlative approaches. They further illustrate the rather impressive 'success stories' of explicit ecosystem manipulations. At the same time the considerable difficulties of designing and operating these experiments for the long time periods that are required are made clear. These problems must be addressed and overcome in order to make continued progress in the development of ecosystem science.

  There appears to be no lack of good ideas on the kinds of experiments that should be performed, and even for the kinds of experiments that should receive priority attention because of the threat of a changing biosphere. A group under the leadership of Kitchell addressed these priorities for aquatic ecosystem experiments and one led by Walker, terrestrial experiments. The former group proposed experiments on whole-system thermal enhancements, toxic waste remobilization, and system trace gas emissions. Further discussions will examine these kinds of experiments specifically for estuaries. The terrestrial group concluded that whole-system experiments on CO₂ enrichment and modifications of precipitation and thermal regimes are most urgently needed. The working group led by Strain address the techniques available for manipulating CO₂ concentrations at differing levels of organization, from a leaf to whole systems. They show that the technology is available for these experiments; indeed there have already been successful whole-system experiments utilizing enriched atmospheric CO₂ concentrations. However, these experiments have been limited to a couple of ecosystem types only. We have the capability, and the need, to increase the scope of these experiments.