Quantifying expert opinion with discrete choice models: Invasive elodea's influence on Alaska salmonids.

Scientific evidence should inform environmental policy, but rapid environmental change brings high ecological uncertainty and associated barriers to the science-management dialogue. Biological invasions of aquatic plants are a worldwide problem with uncertain ecological and economic consequences. We demonstrate that the discrete choice method (DCM) can serve as a structured expert elicitation alternative to quantify expert opinion across a range of possible but uncertain environmental outcomes. DCM is widely applied in the social sciences to better understand and predict human preferences and trade-offs. Here we apply it to Alaska's first submersed invasive aquatic freshwater plant, Elodea spp. (elodea), and its unknown effects on salmonids. While little is known about interactions between elodea and salmonids, ecological research suggests that aquatic plant invasions can have positive and negative, as well as direct and indirect, effects on fish. We use DCM to design hypothetical salmonid habitat scenarios describing elodea's possible effect on critical environmental conditions for salmonids: prey abundance, dissolved oxygen, and vegetation cover. We then observe how experts choose between scenarios that they believe could support persistent salmonid populations in elodea-invaded salmonid habitat. We quantify the relative importance of habitat characteristics that influence expert choice and investigate how experts trade off between habitat characteristics. We take advantage of Bayesian techniques to estimate discrete choice models for individual experts and to simulate expert opinion for specific environmental management situations. We discuss possible applications and advantages of the DCM approach for expert elicitation in the ecological context. We end with methodological questions for future research.

Minimum viable populations: is there a 'magic number' for conservation practitioners?

Establishing species conservation priorities and recovery goals is often enhanced by extinction risk estimates. The need to set goals, even in data-deficient situations, has prompted researchers to ask whether general guidelines could replace individual estimates of extinction risk. To inform conservation policy, recent studies have revived the concept of the minimum viable population (MVP), the population size required to provide some specified probability of persistence for a given period of time. These studies conclude that long-term persistence requires ≥5000 adult individuals, an MVP threshold that is unaffected by taxonomy, life history or environmental conditions. Here, we re-evaluate this suggestion. We find that neither data nor theory supports its general applicability, raising questions about the utility of MVPs for conservation planning.

Design to monitor trend in abundance and presence of American beaver (Castor canadensis) at the national forest scale.

Wildlife conservationists design monitoring programs to assess population dynamics, project future population states, and evaluate the impacts of management actions on populations. Because agency mandates and conservation laws call for monitoring data to elicit management responses, it is imperative to design programs that match the administrative scale for which management decisions are made. We describe a program to monitor population trends in American beaver (Castor canadensis) on the US Department of Agriculture, Black Hills National Forest (BHNF) in southwestern South Dakota and northeastern Wyoming, USA. Beaver have been designated as a management indicator species on the BHNF because of their association with riparian and aquatic habitats and its status as a keystone species. We designed our program to monitor the density of beaver food caches (abundance) within sampling units with beaver and the proportion of sampling units with beavers present at the scale of a national forest. We designated watersheds as sampling units in a stratified random sampling design that we developed based on habitat modeling results. Habitat modeling indicated that the most suitable beaver habitat was near perennial water, near aspen (Populus tremuloides) and willow (Salix spp.), and in low gradient streams at lower elevations. Results from the initial monitoring period in October 2007 allowed us to assess costs and logistical considerations, validate our habitat model, and conduct power analyses to assess whether our sampling design could detect the level of declines in beaver stated in the monitoring objectives. Beaver food caches were located in 20 of 52 sampled watersheds. Monitoring 20 to 25 watersheds with beaver should provide sufficient power to detect 15-40% declines in the beaver food cache index as well as a twofold decline in the odds of beaver being present in watersheds. Indices of abundance, such as the beaver food cache index, provide a practical measure of population status to conduct long-term monitoring across broad landscapes such as national forests.