AVOIDING TRAIN WRECKS IN THE USE OF SCIENCE IN ENVIRONMENTAL PROBLEM SOLVING

Lee E. Benda, N. LeRoy Poff, Christina Tague, Margaret A. Palmer, James Pizzuto, Nancy Bockstael, Scott Cooper, Emily Stanley, Glenn Moglen

______

Lee Benda (email ) is a senior scientist with the Earth Systems Institute, 1314 NE 43rd St., Suite 205, Seattle, WA 98105. LeRoy Poff (email: ) is an Assistant Professor at Colorado State University, Department of Biology, and Ft. Collins, CO 80523. Christina Tague (email: ) is an Assistant Professor at the University of California – San Diego, Department of Geography, San Diego CA 92182. Margaret Palmer () is a Professor at the University of Maryland, Department of Biology, College Park, MD 20742. James Pizzuto (email: ) is a Professor at the University of Delaware, Department of Geology, Newark, DE 19716. Nancy Bockstael (email: ) is a Professor at the University of Maryland, Department of Agricultural & Resource Economics, College Park, MD 20742. Scott Cooper (email: _ is a Professor at the University of California, Department of Ecology, Evolution, and Marine Biology, Santa Barbara, CA 93106. Emily Stanley (email: ) is an Assistant Professor at the University of Wisconsin, Department of Zoology, Madison, WI 53708. Glenn Moglen (email: ) is an Assistant Professor at the University of Maryland, Department of Civil Engineering, College Park, MD 20742.

Abstract: Solving interdisciplinary problems requires, at the onset, an epistemological analysis of the collaborating disciplines. Different forms of knowledge that are available across disciplines are used to identify the scales and resolution of answers that are possible, as well as, those that are not possible and those situations most likely to lead to train wrecks.

INTRODUCTION

Natural ecological systems are being altered in diverse ways and at rapid rates by human actions. Some activities produce sustained impacts on ecosystems that are readily apparent, such as the effects of dam impoundments on river hydrology or the impacts of exotic predators on communities and ecosystems (Mooney and Hobbs 2000). Other anthropogenic influences such as the ecological impacts of global climate change (Kareiva et al. 1993) or the synergistic effects of fishing and habitat alteration on fishery stocks are more difficult to identify and involve many interacting factors. Finding solutions to environmental problems is challenging because it requires an understanding of complex scientific issues as well as how socio-political factors act to dictate feasible or desirable courses of action.

Because of society’s interest in stemming environmental degradation, knowledge from a range of scientific disciplines must often be focused on environmental problems. Completing interdisciplinary analyses, however, is difficult for numerous reasons. A fundamental problem is that science is inexorably shaped by the questions that it addresses and by the scientific and socio-political context that determine these questions (Kuhn 1970, Latour 1993, Pickett et al. 1994). Hence, existing scientific theories and technologies may originate from questions that are unrelated to today's interdisciplinary and complex environmental problems. This may lead to a shortfall between the concepts, theories, analytical methods, and empirical relations available to explainand predict phenomena and the knowledge and methodological requirements needed to answer environmental questions (Ford 2000, NSB 2000, and many others). Moreover, it also can be difficult to find a common language so that a diverse set of disciplinarians can communicate (Wear 1999, Sarewitz et al. 2000).

In principle, the participation of individuals from a diverse set of disciplines should enhance the success of problem solving, and it often does so. However, addressing problems with multiple disciplines also may create difficulties owing to mismatches in space and time scales, in forms of knowledge (e.g., qualitative vs. quantitative), and in levels of resolution and accuracy addressed by different fields (e.g., Herrick 2000). These difficulties can lead to train wrecks[1] defined here as failures to solve environmental problems because of an inability to effectively communicate among scientists, regulators, and the public, because current scientific understanding or predictive capacity is inadequate in some fields, and/or because of disagreements about problem solving strategies. We also recognize that there are other reasons for train wrecks, often brought about by political realities or conflicts between private interests and the interests of society, but we limit our analysis in this paper to scientific reasons. The limitations and opportunities that arise from the diversity of scientific training, approaches, and knowledge collectively embodied by an interdisciplinary team are often misunderstood by the public, policy makers, or even by scientific practitioners themselves.

Our goal in this paper is to propose a process for increasing the success of interdisciplinary collaborations. The process begins with an awareness of the epistemology of various disciplines, including an understanding of the origin, nature, methods, and limits of knowledge. An epistemological analysis is used to identify multiple pathways by which knowledge and approaches contributed by a diverse set of disciplines can be applied to environmental questions, including identifying possible solutions and the potential for train wrecks. Our general approach can be applied to a broad range of interdisciplinary questions; however, we illustrate its use by focusing on one general class of environmental questions, i.e. the effects of land use changes on riverine ecosystems.

CREATING AN INTERDISCIPLINARY SOLUTION SPACE

All solutions to environmental problems begin with questions. When examining the effects of land use changes on riverine ecosystems, we can pose specific questions: Can we predict the effects of urbanization or logging on flow regimes and erosion, and how do these factors, in turn, affect aquatic biota? Answers to this question will require input from disciplines as diverse as economics, surface water hydrology, geomorphology, and riverine ecology. These disciplines are hierarchically arranged to depict a cascading sequence of hypothesized causes and effects (or inputs and outputs), representing linkages between the questions and knowledge of different disciplines (Figure 1). Thus, economic factors will drive urbanization or logging, and economists can forecast spatial and temporal patterns in future land uses in a watershed. Hydrologists and geomorphologists, in turn, use this information to predict how the storage and flux of water, sediment and nutrients will change in stream channel networks draining the watershed. Predicted changes in physical and chemical conditions in streams are then used to evaluate the possible ecological consequences of the forecasted land use changes.

Central to the application of interdisciplinary approaches to successful problem solving efforts is the recognition that each discipline produces knowledge (theories, models, statistical relations, empirical descriptions) that has a particular history and structure. The knowledge structures of disciplines (e.g., economics, surface water hydrology, geomorphology, and riverine ecology) include the spatial and temporal scales over which the knowledge applies, the qualitative vs. quantitative nature of knowledge, and the levels of resolution and accuracy in predictions. By mapping each discipline’s forms of knowledge on axes related to their space and time scales, resolution, and accuracy, and by superimposing and linking the areas of knowledge produced by each discipline, we produce a “solution space” (Figure 1). The solution space is used to identify the feasible scales, accuracy, and resolution of solutions produced by an interdisciplinary effort. We propose that such a solution space, structured around the epistemological analysis of collaborating disciplines, would reveal a finite set of “pathways” that lead from environmental questions to answers that identify both opportunities and potential train wrecks of scientific investigations and applications.

Many attempts to answer interdisciplinary questions fail because participants assume that there is a particular pathway to an answer at the outset, e.g., quantitative data will be used to build quantitative models which predict outcomes with a high degree of precision. In many cases, however, multiple solution pathways may be possible, depending on the exact question and the knowledge structures of the disciplines applied to the problem. We refer to these multiple answers to environmental questions as “question – knowledge pathways” (Figure 2), and contend that explicit recognition of these pathways would help avoid train wrecks in interdisciplinary research or problem solving. The explicit recognition of the histories and knowledge structures of different disciplines should advance interdisciplinary problem-solving efforts.

Limitations to interdisciplinary problem solving include the absence of critical pieces of scientific knowledge or a question addressing dimensional scales that lie outside of the available forms of knowledge (pathways 1, 2, 3, and 4, Figure 2). Opportunities created by interdisciplinary problem-solving may include: 1) altering the environmental question to more closely match the available forms of knowledge, which may require omitting certain types of questions; 2) legitimizing qualitative, statistical, and other, new forms of knowledge (Benda et al. 1998; Gaff 2000); and 3) identifying research or data needed to answer environmental questions (pathways 5, 6, and 7; Figure 2). Acceptance of one of these “opportunities” may save the problem-solving effort from failure.

APPLYING INTERDISCIPLINARY KNOWLEDGE TO ENVIRONMENTAL QUESTIONS

The process that we used to construct the interdisciplinary solution space (Figure 1) and to identify the pathways leading from environmental questions to answers (Figure 2) is described in the remainder of this paper. Delineating and interpreting the solution space requires identifying the knowledge structure of each relevant discipline, focusing on the origins (i.e., history of questions and approaches), methods, and limits of knowledge of each discipline. To begin, we map the spatial and temporal scales addressed by the various forms of knowledge in each of the four disciplines relevant to examining land use impacts on riverine ecosystems on a Stommel diagram (Stommel 1963) (Figure 3). The diagrams in Figure 3 are hierarchically arrayed in the solution space in Figure 1.

The epistemological analysis recognizes that the history of a discipline strongly influences its forms of knowledge (Kuhn 1970; Schaffer 1996), indicating that the knowledge base of a discipline depends, in part, on the prevailing social and scientific climate during the time the knowledge was generated (Feyerabend 1978). The research traditions embodied in different disciplines often dictate what techniques are available, what variables are measured, and how data are analyzed and interpreted. Some detail on the traditions and bodies of knowledge in the four disciplines as they pertain to our focal land use question are presented in Boxes 1- 4. Comparison of this information reveals how knowledge incompatibilities may arise when disciplines interact. For example, environmental economics has developed as a sub-discipline of economics during the last few decades in direct response to the fact that environmental amenities and ecological functions are valuable to people individually and to society as a whole, but can not be traded on markets like other goods. This had led to analysis of the effects of land uses on environmental goods and services, including air and water quality and landscape aesthetics. The development of riverine ecology also has accelerated since the 1970s during a time of increasing environmental awareness. In contrast, the development of hydrology and, to some extent, geomorphology was driven by practical engineering concerns in the late 19th and early to mid 20th centuries related to the protection of human structures from floods and landslides, and the construction of stable, artificial waterways. During that period and up to the present, geomorphology was also focused on landform evolution. Until recently, hydrology and geomorphology were not directly concerned with environmental issues.

Commonly, resource managers, regulators, and policy makers request scientific predictions that are accurate and precise for large systems (i.e., large watersheds). The expectation of precise, quantitative predictions arises from the predictive successes of quantitative approaches in other scientific disciplines, such as chemistry and Newtonian physics (Gell-Mann 1994). For instance, there is presently great concern about how spatial patterns of urbanization will affect stream ecosystems (i.e., "Smart Growth" issues, Palmer et al. accepted). This expectation often exposes one of the most common sources of incompatibilities among disciplines - differences in spatial and temporal scales, a problem that can be recognized and possibly circumvented by use of a solution space (Figure 1). For example, many economic theories about the spatial arrangement of land use (e.g. Alonso 1964; Krugman 1996) characterize in a qualitative way the expected nature of land use pattern around city centers. They are broadly applicable to urban/suburban centers throughout the U.S. over decadal time horizons, but provide little spatial detail and no precise quantitative predictions. In contrast, empirical economic research that models spatially explicit land use change with an aim of high levels of accuracy (e.g. Bockstael 1996; Landis 1995) is typically organized around small-scale questions (Box 1, Figure 3), effectively precluding forecasting over periods of decades and over broad regions. Even more critical, the spatial scales used in economics are at the level of either neighborhoods or a land market as characterized by a city center and surrounding suburbs. The relevant geographic scale for detailed economic study is generally not compatible with the spatial scales that govern the function of riverine ecosystems (entire large watersheds).

Spatially distributed hydrological models (Box 2, Figure 3) can be used to predict relationships between land use patterns and flow for some small catchments, but data limitations (i.e., precipitation and flow records) typically preclude the application of these models to all the small streams draining a large basin. In large watersheds, fine-scale variation in the hydrological behavior often can be ignored when estimating hydrological responses to land use changes in the entire basin. Flow regimes at the outlet of the large basin may be estimated with reasonable accuracy using relationships between total discharge and average characteristics of the entire basin. These lumped approaches do not provide fine-scale information on watershed responses and they do not incorporate information on complex, continuous land use changes or the spatial distribution of these changes (e.g., riparian vs. upland changes) (Moglen and Beighley submitted).

Geomorphology, like other areas of geology, examines natural systems that span very large time and space scales (regional landscapes over millennia) where current environmental conditions are strongly contingent on past events (i.e., geologic and climatic history). As a result, geomorphic predictions are often imprecise, qualitative, and have low accuracy compared to economics or to hydrology. Moreover, because each geomorphic situation is somewhat unique and the large scale of focus precludes experimental manipulations, opportunities for repeating observations and falsifying hypotheses are few (Schumm 1991). In quantitative geomorphology, the precision of predictions generally declines with increasing scale and the number of model parameters. For example, geomorphological theories and concepts encompass a broad range of scales (Box 3, Figure 3) and can be used to address questions at both small and large scales (i.e., sediment transport over meters during a single storm or in channel networks over decades). However, theories and models addressing sediment transport produce predictions that have low accuracy, even at small scales (Gomez and Church 1987). If the effects of land use changes on sediment supply and transport at large scales are of interest, then the number of parameters needed to predict these changes rises rapidly (hillslope erosion and attendant changes in channel morphology must be considered). Although this scale of analysis can lead to valuable insights into whether small or large changes in channel state are anticipated (i.e., direction and general magnitude, e.g., Schumm 1977), the accuracy of predictions at small scales is quite low (e.g., channel scour in reaches during individual storms).

Finally, the ecological knowledge required to address the effects of land use changes on riverine ecosystems encompasses a large number of elements and parameters (Box 4, Figure 3). The complexity of ecological systems include diversity of food web components, linkages between them, influences of multiple environmental factors, and the contingent role of history. Ecological patterns and processes are shaped by present and past environmental conditions and by interactions among species within their changing environments. The role of past conditions or events on river systems is difficult to quantify; however, the impacts of historical conditions on current patterns can be substantial (e.g., Sedell and Froggatt 1984). The great variation in environmental conditions and species composition across riverine landscapes contributes to large spatial and temporal variation in the structure and function of ecological systems. Consequently, almost all of the general principles used in riverine ecology are conceptual and qualitative (i.e., River Continuum Concept, Flood Pulse Concept, etc.) (Box 4, Figure 3).

Within qualitative, conceptual frameworks, however, riverine ecologists often use quantitative, analytical approaches to express relationships between ecological variables and environmental conditions. Both statistical and numerical modeling approaches are now used widely in riverine ecology, yet their ability to make precise quantitative predictions about ecological responses to land use change is often limited. Quantitative predictions of land use effects on a riverine system require reach-scale data collected under particular conditions. Usually, however, extrapolating local models to other locations (where detailed data are lacking) is problematic, owing to the increased complexity of larger systems and the role of history in producing spatially unique patterns. On the other hand, statistical models can be developed for large scales to express important relationships between dominant environmental variables at those scales and ecological measures.

So-called black box models can be used to assess the probability and magnitude of change at an individual site; however, the predicted outcomes are typically contingent on many assumptions (about history; the validity of extrapolations to other sites, times, or scales; the effects of unmeasured environmental variables; differences between large-scale average vs. local responses). If these assumptions are violated, then quantitative predictions become difficult and often inaccurate (Palmer et al. submitted). Of course, limitations imposed by the effects of historical contingency and local conditions on outcomes apply to many scientific disciplines (e.g., Ulanowicz 1997). Given these considerations, riverine responses to land use change will often be couched as qualitative predictions. For example, ecologists may be able to say that pollution tolerant taxa will increase and sensitive taxa will decline if a forested watershed is developed; however, accurately predicting the change in abundance for individual species or the change in community diversity is not feasible.