© 2002 by CRC Press LLC CHAPTER 14 Enhancing the Use of Ecological Models in Environmental Decision-Making Lev R. Ginzburg and H. Resit Akçakaya Although ecological models have been used to better understand natural systems to support envi- ronmental management (e.g., Holling 1978; Hilborne and Mangel 1997; Walters 1986; Breininger et al. in press), relatively few applications to toxic chemical issues exist. One of the most important factors preventing widespread use of models in decision-making is a lack of modeling training, which prevents many managers from determining the best kind of model and scale of resolution for solving a given problem (Ginzburg and Akçakaya in press; Breininger et al. in press). Another important factor is the lack of appropriate and relevant data (Ginzburg and Akçakaya in press). Moreover, untrained managers often do not know what types of data to collect to optimize model use. As shown in our review of models above, many ecological models already exist that can be applied to practical problems in environmental management of toxic chemicals. Ginzburg and Akçakaya (in press) identified four different areas that require additional investment to enhance the use of ecological modeling in decision-making: training, applying existing models, integrating existing models, and developing new, case-specific models. TRAINING AND EDUCATION Teaching environmental managers how to use existing models is the most efficient way to enhance model use in the short term. Jackson et al. (2000) provide an excellent introduction to the practice of ecological modeling that is suitable for any audience. However, the specific content and format of training courses should be based on the audience. Audiences for such training courses may be divided into two general categories: environmental managers and technical support personnel. The first potential audience for training courses includes environmental managers — manage- ment and research decision-makers in government agencies and industry. Managers often are unfamiliar with the potential of modeling. Topics for educating managers in the use of ecological models include: 1574CH14.fm Page 205 Tuesday, November 26, 2002 6:28 PM © 2002 by CRC Press LLC • The types of questions that can be addressed by using ecological models • Selecting the appropriate model • The types of data needed for different types of models • Interpreting the results of models • Modeling, interpreting, and communicating risk and uncertainty • Identifying inappropriate use of models • Examples of successful applications of models Technical personnel are another audience who may be developing or reviewing models. To judge the technical merits of models, they need to have a basic understanding of the fundamentals of modeling. Topics for technical personnel should include all of the topics for managers, as well as the following topics for more advanced researchers and technical personnel: • Components of different types of ecological models • Trade-offs between complexity (realism) and practicality (data availability) • Collecting the appropriate data • Analyzing data to estimate model parameters • Incorporating variability and uncertainty • Presenting model results APPLYING EXISTING ECOLOGICAL MODELS The simplest way to develop a model is to apply an existing model to a management question. “Applying” means estimating the parameters of the model from data that are specific to the location, species, and system in question and using the model to address a specific management question. Many existing models have been successfully used to guide management decisions (Holling 1978; Walters 1986; Bartell et al. 1999). Population and metapopulation models have been successfully applied to many environmental issues, especially questions regarding recovery of threatened and endangered species and manage - ment of harvesting. For example, in May 1995, the Oregon Department of Fish and Wildlife used an age-structured model to report on the biological status of the marbled murrelet (Brachyramphus marmoratus) in response to a petition to list the species under Oregon’s Endangered Species Act (Oregon Department of Fish and Wildlife 1995). This species was later listed as threatened under the act. In another case, a metapopulation model was used to evaluate the effectiveness of translo - cation as a management tool for the endangered helmeted honeyeater (Lichenostomus melanops cassidix) and to support the decision regarding the timing of the release of this species into its new environment (Akçakaya et al. 1995). Data from a GIS and a metapopulation model were used to determine viable population size for the Florida scrub jay (Aphelocoma coerulescens) (Root 1998) in the context of four reserve designs developed in a Habitat Conservation Planning process for nonfederal scrub habitat in Brevard County, Florida (Brevard County Office of Natural Resources 1995). Another metapopulation model was applied to the redhorse (Moxostoma sp.) populations in the Muskingum River in Ohio (Root et al. 1997) to model the thermal impact of a proposed increase in power plant operation, which was later approved by the Ohio EPA. Many existing ecological models can be applied to support or guide management decisions. Such applications require collection and statistical analysis of site specific data to estimate model parameters. Once the model parameters (and their uncertainties from measurement error and natural variability) have been determined, the application of an existing model requires minimal research effort. Therefore, the primary scientific issues in the application of existing models involve data analysis methods. These include survival estimation methods based on mark–recapture data, methods for estimating spatial, temporal, and error variance components, and variance due to age and sex. Applications to toxic chemical issues require use of available toxicity data, preferably 1574CH14.fm Page 206 Tuesday, November 26, 2002 6:28 PM © 2002 by CRC Press LLC dose– or concentration–response curves, or the collection of appropriate toxicity data. Most ecological models that do not already have functions for incorporating toxic chemical effects can be easily modified to account for such effects. Often, constraints on the immediate use of ecological models to address toxic chemical problems will arise from lack of toxicity data, not from lack of an appropriate model. Table 14.1 specifies the expected effort (in time) and expenditure (in thousands of dollars) required to use models from the broad categories for a specific application. For each model category, we assume that collection of field data would not be required. Also, at least one model within the category must exist that is appropriate for the task without requiring substantial research or major modifications for model development. All effort and cost figures are approximate and include compilation and analysis of data from the literature, and parameterization, calibration, and sensi - tivity analysis of the model for a single toxic chemical. INTEGRATING EXISTING MODELS In the recent past, model development has involved integrating existing models — for example, habitat models and demographic models or life-history matrix models and spatial models (e.g., Akçakaya 1998a,b). One of several such potential developments involves linking simple (scalar) population models to models based on allometric relationships. The result can be used in screening assessments with minimal or no field data. An important related research issue is testing whether the level of conservatism (precaution) of this approach is comparable with the level required in a screening test. Another type of integration links fate-and-transport models to ecological models. Although this approach has been used in specific cases, a general modeling platform is needed to link physi - cal/chemical models (e.g., hydrological models), dose–response models, and population or meta- population models (Reinert et al. 1998; Ginzburg and Akçakaya in press). General models are also needed to integrate food-web and metapopulation models. Such inte- gration would allow the modeling of trophic interactions in a spatially structured habitat with different metapopulation structures for different species. With this approach, coordinating temporal and spatial scales and resolutions is an important research issue. Table 14.1 Estimated Effort and Expenditure Required for Application of Ecological Models a Model Category Effort Required (in time) Required Expenditure (in $1000s) Toxicity-Extrapolation models 0.2–2 months 2–40 Population models — scalar abundance 0.5–2 months 10–40 Population models — life history 1–3 months 20–80 Population models — individual-based (with software) 2–5 months 40–100 Population models — individual-based (without software) 0.5–1.5 years 100–200 Population models — metapopulation 1–3 months 20–80 Ecosystem models — food web 2–4 months 40–100 Ecosystem models — aquatic 0.3–1 year 40–200 Ecosystem models — terrestrial 0.3–1 year 40–200 Landscape models — aquatic and terrestrial 0.5–2 years 60–500 a The estimates of effort and expenditure shown in the table depend on the assumption that the ecological model is run to assess the effects of a single chemical. They do not encompass the toxicity assessment, which varies depending on the number and types of chemicals addressed. The toxicity assessment and derivation of dose– or concentration–response curves may require effort and expenditure on the order of that shown in the table for applying the ecological models. This table does not address collection of field data or other parts of an ecological risk assessment (e.g., problem formulation, exposure assessment, and risk characterization). 1574CH14.fm Page 207 Tuesday, November 26, 2002 6:28 PM © 2002 by CRC Press LLC Temporal and spatial scales are also important in linking metapopulation and landscape models. Incorporating landscape dynamics in the spatial structure of metapopulation models allows the effects of landscape management options on the viability of key species to be evaluated. Integration may also involve several models of the same type. For example, linking population models for a set of several target species allows decisions to be made in a multispecies context without constructing a complex food-web or ecosystem model. DEVELOPING NEW, CASE-SPECIFIC MODELS In most cases, an existing model can be used to address management questions. However, many cases require developing a new model. Such a requirement can often be met by integrating existing models but, in rare cases, the management question may require a completely new modeling approach. INVESTMENT TRADE-OFFS Investments in enhancing the use of ecological models in environmental decision-making involve three trade-offs represented by the diamond-shaped boxes in Figure 14.1 (Ginzburg and Akçakaya in press). The first trade-off is between research and training. Of these two options, investment in training yields the greatest short-term return. For managers, a one-day workshop is the most suitable approach. For technical staff and researchers, the most cost-efficient approach is Internet-based teaching supplemented by telephone. The second trade-off is between model development and application (Figure 14.1). The best option depends on the time constraints for management decisions. Both options require collecting and analyzing data, after which minimal additional investment of time and research effort is required for application. Therefore, for short to medium time-horizons, the most efficient way to develop models is to use existing models with the data needed to answer the specific question at hand. In most cases, the availability of data, and not the model, is the limiting factor. The third trade-off is between integrating existing models and creating new models (Fig- ure 14.1). Integrating existing models results in models with enhanced capabilities in the midterm, whereas creating new models results in new models in the longterm. The resources required to develop new models depend on the type of model and the population, ecosystem, or landscape being modeled. As investment resources increase for enhancing the use of ecological modeling in environmental decision-making, the type of activity emphasized should shift from training to model application and development (Figure 14.2). Shifts in activity as a result of changes in funding targets also correspond to a sequence in these activities over time. With small investments, funding should focus on training of managers and technical staff. With improved knowledge as a result of training, increasing invest - ments can be focused on application of existing models. Over time, the increased knowledge from applying models allows integration of existing models and development of new models. 1574CH14.fm Page 208 Tuesday, November 26, 2002 6:28 PM © 2002 by CRC Press LLC Figure 14.1 Series of three decisions for making research investments to enhance the use of ecological models in environmental decision-making. Note: Diamond-shaped box indicates a decision point. (From Ginzburg, L.R. and H.R. Akçakaya (in press). Science and management investments needed to enhance the use of ecological modeling in decision-making. In V. Dale, Ed., Ecological Modeling for Environmental Management, Springer-Verlag, New York. Redrawn with permission.) Figure 14.2 Efficient investment approach for enhancing the use of ecological models in environmental deci- sion-making. (From Ginzburg, L.R. and H.R. Akçakaya (in press). Science and management investments needed to enhance the use of ecological modeling in decision-making. In V. Dale, Ed., Ecological Modeling for Environmental Management, Springer-Verlag, New York. Redrawn with permission.) 1574CH14.fm Page 209 Tuesday, November 26, 2002 6:28 PM . additional investment to enhance the use of ecological modeling in decision-making: training, applying existing models, integrating existing models, and developing new, case-specific models. TRAINING. INVESTMENT TRADE-OFFS Investments in enhancing the use of ecological models in environmental decision-making involve three trade-offs represented by the diamond-shaped boxes in Figure 14. 1 (Ginzburg. limiting factor. The third trade-off is between integrating existing models and creating new models (Fig- ure 14. 1). Integrating existing models results in models with enhanced capabilities in