Northern Prairie Wildlife Research Center
GIS data bases
The term geographical information system (GIS) is currently in vogue, but the idea that spatial variation in the landscape is a key to informed management decisions has been around since 1969, when McHarg (1992) first published 'Design with Nature'. McHarg understood the complexity of landscapes and proposed a methodology, GIS, that had the potential to 'put Humpty Dumpty back together again' (American Society for Photogrammetry and Remote Sensing, 1988). The power of GIS has been greatly expanded by advances in remote sensing and computer technology. This tool is valuable for analyses of landscape ecology (e.g. Turner, 1989) and for monitoring natural resources. There have been numerous excellent applications of GIS as a tool in decision making, but many of these examples are from local areas. Unfortunately, there are few examples where the methodology has been used on a scale as large as the Prairie Pothole Region. The problem lies not with any shortcoming of GIS nor with the availability of hardware systems and GIS software. Rather, it results from a lack of appropriate data being available to the user. Data are the most important and most expensive components of a GIS.
The first step in developing a GIS is not acquiring hardware and software, but specifying the purpose for the GIS. Too often, agencies have acquired a GIS facility, only then to ask 'What do we do with it?' The questions to be answered with a GIS determine the type, quantity, and quality of the data required.
Data gathered by satellites such as the Landsats are generally available, but costs may be prohibitive for an area as large as the Prairie Pothole Region. Koeln et al. (1988) demonstrated that satellite data can be used as a planning tool in waterfowl management. Many habitat management questions for the Prairie Pothole Region involve wetlands and associated wildlife. Fortunately, detailed wetland mapping of the region by the US Fish and Wildlife Service's National Wetland Inventory is nearly complete (Wilen, 1990) . These maps are being converted to digital products, but the task is not yet finished. Digital data for political jurisdictions and cultural and hydrological features are available from the Earth Sciences Information Center of the US Geological Survey's National Mapping Program.
For many questions, interpretation may require data from one or more specific time periods. Obtaining data for the entire region may be impractical, but sampling can greatly reduce the need for data, with GIS applied to sample plots. Cowardin et al. (1988a) used GIS techniques applied to sample plots in conjunction with the mallard productivity model described below.
Gap analysis is a tool developed by the US Fish and Wildlife Service and cooperators (Scott et al., 1993) . It is intended to determine how much biological diversity is in the current system of protected natural areas. Diversity is indicated indirectly by surrogates such as potential distributions of vegetation, vertebrates, and butterflies. Geographical information systems are used extensively to consolidate the information. Gaps are unprotected (privately owned) areas potentially rich in biological diversity. The premise behind gap analysis is that crises in populations, such as endangerment, can be avoided by proactive identification of areas most suitable for preservation. One deficiency of gap analysis is that, by virtue of its extensive nature, the results are too crude for use in local management decisions and do not provide an adequate measure of diversity for grassland habitat. It identifies areas of potential occupancy, rather than actual distributions of species. Also, it does not address the population viability of animals or plants that might be in an area. Its strength is in consolidating information and providing a vivid portrayal of areas bearing further scrutiny.
Habitat suitability index models
Habitat suitability index (HSI) models represent another effort by the US Fish and Wildlife Service to develop a tool for use by field managers. Wildlife managers often need to predict the impacts of human activities on certain animal populations. Typically such predictions are made subjectively. The Service attempted to develop methods for objective and quantitative assessment. This process, involving habitat evaluation procedures, used HSI models for certain 'indicator wildlife species' in affected habitats. The premise was that certain habitat features, which can be measured with relative ease, could be used to estimate the carrying capacity of the habitat for the species of interest (US Fish and Wildlife Service, 1980, 1981). Most HSI models were developed through the use of expert opinion. The few tests that have been conducted have found limited correspondence between predictions of the model and actual numbers of animals, especially for mobile species inhabiting complex habitats (e.g. Johnson et al., 1989 and references therein). Nonetheless, such models do provide objective criteria for assessing environmental impacts.
The mallard model
Simulation models are useful to aid decision-making in habitat management; some incorporate information on both landscape features and population dynamics. An example is the model of mallard productivity developed at the Northern Prairie Wildlife Research Center (Johnson et al., 1987). It is distinctive in that it links population processes to landscape attributes (Cowardin et al., 1988b) . It has aided management decisions in the US Fish and Wildlife Service's Regions 3 and 6 and in planning for the North American Waterfowl Management Plan in both the United States and Canada. A regression component predicts breeding population from the amount and type of wetland habitat available. Inputs to the productivity model include availability, attractiveness, and hatch rates of nests for classes of nesting habitat. These are combined to produce an estimate of the recruitment rate, which then can be applied to the breeding population to estimate the number of young produced. Various applications employed habitat data from a sample of more than 500 4-square-mile areas to predict the production of mallards for the entire Prairie Pothole Region. For site-specific applications, such as Hamden Slough National Wildlife Refuge, a habitat inventory was completed to furnish the habitat availability data. The model can be manipulated to simulate various treatments and combinations of treatments. The model-based procedure gives managers some indication of the possible outcome of proposed management before actual application.
The use of such a model is not without hazard; results should not be blindly accepted. Models cannot tell you what to do, but they can be helpful in evaluating alternatives. Nor can models, by themselves, set priorities. Even though the predictions derived from the model may not represent exactly what will happen, they force some essential actions in habitat management: (1) assembling the best available data; (2) understanding where data are inadequate and acknowledging the importance of assumptions that must be made; (3) gaining insight into potential outcomes of a number of management scenarios prior to actual application; (4) evaluating the feasibility of objectives; (5) aiding the assignment of resources.
The approaches mentionedgap analysis, HSI modeling, and the mallard productivity modelrepresent points along a continuum of biological detail and resolution. At one extreme, gap analysis provides a perspective of the potential habitat. In the middle, HSI and similar models offer an assessment of the actual habitat and potential population. At the other extreme, the mallard model estimates not only the actual population, but also its productivity.