13.1 INTRODUCTION

Water is a fundamental societal need. Global climate change will affect both the quantity and quality of water resources. Although the specific features of rates and magnitude remain the subject of continuing debate, all reasonable forecasts of global climatic response to increases in atmospheric carbon dioxide and other greenhouse gases include increases in mean temperatures at middle latitudes and a substantial increase in mean sea level (NRC, 1986). Coastal systems and the regional hydrologic cycles of continents will be profoundly altered. While there is no doubt that the riverine conduits and oceanic recipients of a changing hydrologic cycle will be responsive to global change, we have focused our attention on the systems we deem most sensitive and amenable to the use of whole-system experimentation. With these as givens, our group attempted a collective vision of an unknown future. We constructed a consensus of the ecosystem-scale experiments that would offer most powerful insights to interests of both the research and management communities.

   Ecosystem experiments resolve controversy and rapidly achieve insights that would take far longer through observational and/or small-scale experimental studies (Likens, 1985). The limiting nutrient controversy in limnology is one example of the power of ecosystem experiments to reduce disagreement among scientists and prompt effective management action (Schindler,1988). By 1970, several decades of debate deriving from correlational and small-scale experimental studies still could not resolve the relative importance of carbon, nitrogen, and phosphorus in controlling water quality of eutrophic lakes (Likens, 1972). Whole-lake enrichment experiments showed clearly that phosphorus was the critical nutrient. Phosphorus control is now the central element of water quality management (Schindler, 1988).

   Proper scaling in space and time is an essential requisite in the design of ecosystem experiments. If the areal or temporal dimensions are insufficient, results are likely to be confusing or misleading (Carpenter and Kitchell, 1988). Statistical issues are subordinate to scale considerations as the history of statistics derives from subsampling and replicated test plots rather than the holistic view of entire systems. Although ecosystem experiments often will not meet the replication requirements of conventional statistics, the need for effective research and management at the whole-system level demands that we overcome the constraints of statistical tradition (Hurlbert, 1984). Fortunately, new statistical and analytical tools (Walters, 1986; Carpenter et al.,1989) provide a rigorous foundation for detecting change in experimentally manipulated systems.

   Ecosystem experiments remain rare despite their large, positive impact on the progress of science and development of management practices. Lack of access to research sites of sufficient size and institutional constraints are the main limitations. The prospects of global climate change demand that our society effectively plan for and respond to extraordinary changes in our water resources. We must accept the responsibility for extraordinary commitments of resources and research talent in order to gain the powerful insights of ecosystem experimentation.

   In this report we emphasize scientific problems requiring the scale and understanding that can be obtained only by the power of whole-ecosystem experiments. We emphasize those types of systems and specific issues where we feel societal interests will be best served by immediate attention. Assessments and recommendations are presented for discrete system types but include the assumption that many important changes in aquatic systems will derive from changes in the terrestrial systems located 'upstream' in the hydrologic cycle. In addition, we point to certain major research questions that may be better suited to conventional methods such as comparative or survey approaches, incisive and well-replicated experimental designs in microcosm or mesocosm systems, process rate responses and modeling.