ORAL PRESENTATIONS

Population Models for Manatee Conservation in Florida. BRUCE B. ACKERMAN, Florida Marine Research Institute, Florida Dept. Environmental Protection, St. Petersburg, FL.

Recent data on the endangered Florida manatee (Trichechus manatus) yield controversial information on their population dynamics. Field data are ambiguous about the population trends. Two population models were compared for 5 regional subpopulations. A PVA model (Vortex 6.3) used stage-specific survival and reproduction from long-term telemetry and photo-identification studies. An independent discrete equation model used regional numbers of carcass recoveries, reproductive rates from carcasses, and aerial counts. Projections from the 2 models were validated compared to aerial survey and carcass trends. Results were compared to PVA and Lotka models by previous researchers. Models suggest the populations have increased in the 5 regions, but this trend could reverse as human impacts increase. Both the annual number of carcasses recovered (3270 carcasses, 1974-97) and aerial counts have increased. The statewide aerial count of 2639 manatees in February 1996 exceeded previous counts. But carcass numbers in 1996 were also extremely high (n=416 deaths) following catastrophic mortality from a cold winter (47) and from red tide (149). Long-term aerial studies suggest increases of 1 to 10%/year in some areas. Reconciling the divergent data has been controversial. Mortality varies among regions because of differing densities of manatees, humans, and watercraft, crowded waterways, etc. Recovery efforts focus on reducing human-related deaths and injuries, especially watercraft collisions. These models help reconcile differences in field data, assess current population dynamics, predict future trends from increasing human impacts, and improve conservation strategies.

Making PVAs Useful Despite Uncertainties. H. RESIT AKÇAKAYA, Applied Biomathematics, Setauket, NY.

Parameters and structure of models used in PVAs are often estimated based on insufficient data, resulting in large uncertainties. PVAs should treat these uncertainties as well as natural variability in an appropriate way. Stochastic demographic models can be used to systematically incorporate uncertainties due to lack of information or measurement error, in addition to variabilities due to environmental fluctuations and demographic stochasticity. When PVAs are used for impact assessment or making management decisions, accounting for these two fundamentally different types of variability allows the models to be more sensitive to the impact or the management action. Results of recent habitat-based PVAs (involving the Red-cockaded Woodpecker and the Northern Spotted Owl) were sensitive to the effects of alternative management actions or human impact, despite considerable model uncertainty. These results suggest that in most circumstances it is possible to make robust assessments and management decisions despite the inevitable uncertainties. One of the conditions for this involves decomposing the uncertainties when making assessments and making the comparisons with respect to key model assumptions. Other conditions include choosing a model with the appropriate level of detail, using probabilistic results and relative risks, focusing on risk of decline (instead of risk of extinction), and choosing the appropriate time horizon. When these conditions are met, PVAs can be used to compare and rank management alternatives in terms of their effect on the viability of the species studied.

The Role of Genetics in Population Viability Analysis. FRED W. ALLENDORF*, Div. Of Biol. Sci., Univ. Montana, Missoula, MT, and NILS RYMAN, Div. Of Pop. Genet., Univ. Stockholm, Sweden.

Controversy over the importance of genetic considerations in Population Viability Analysis (PVA) in the last decade has occurred partially because of confounding of the different roles that genetics may play in performing PVA. Genetics can play at least three distinct roles. The first is to provide an understanding of the demographic and reproductive relationships of the species to be modeled. For example, do the fish present in a lake comprise one or many separate demographic units? Molecular genetic analysis can be used to detect multiple reproductively isolated demographic units within one geographical area. Genetic analysis may also provide important information about demographic and reproductive relationships within a single local population (e.g., variance in reproductive success, effective population size, or amount of selfing). Second, the inbreeding effect and loss of genetic variation within small populations may decrease the probability of persistence of isolated populations. The consideration of such effects may be important in accurately predicting the viability of some populations. Third, in some circumstances it is important to use genetic goals along with PVA. Persistence of a population for some specified period of time (say 100 or 200 years) is not a sufficient goal if erosion of genetic variation occurs that threatens the longer term persistence of a population. Our presentation will suggest approaches to integrate genetics into PVA in all three of these roles.

Formal Inference from a Suite of PVA Models. DAVID R. ANDERSON*, Colorado Cooperative Fish and Wildlife Research Unit, USGS, Biological Resources Division. Ft. Collins, CO, and KENNETH P. BURNHAM, Colorado Cooperative Fish and Wildlife Research Unit, USGS, Biological Resources Division, Ft. Collins, CO.

Predictions of population viability are typically model-dependent. We provide a theory, based on Kullback-Leibler information, for making formal inference from more than one model (i. e., multi-model inference). We assume (1) relevant quantitative data, (2) a set of a priori models carefully derived and supported, and (3) either likelihood or least squares estimation of model parameters. Quantities can be easily computed to allow models to be ranked from best to worst and scaled to indicate relative "distances" between the models. A transformation allows estimation of the (discrete) likelihood of model i, given the data (i.e., L (model i | data)) and normalizing these to represent probabilities. A confidence set for models can be constructed. Inference concerning predicted values can be based on all the models in the a priori set, using "Akaike weights". Finally, estimates of precision can be computed, including a variance component for model selection uncertainty. Information-theoretic methods avoid the problems surrounding null hypothesis testing (e. g., the arbitrary a -level, the interpretation of P-values, the arbitrary classification of results into biological meaningless classes "significant’ and "not significant"). These issues are important in both applied and theoretical settings.

Utility of Population Viability Analyses for Recovering Kirtland's Warblers. JONATHAN BART*, USGS Forest and Rangeland Ecosystem Science Center, Snake River Field Station, Boise, ID, and CAROL BOCETTI, USGS Patuxent Wildlife Research Center, Laurel, MD.

We evaluate a series of population viability models to determine which—if any—provide reliable, new guidance on the best ways to recover an endangered species, the Kirtland's Warbler. The Kirtland's Warbler Recovery Team has defined "recovery" as occurring when population size reaches 1,000 pairs, and the recovery program, which involves creation of new habitat (jack pine) each year, is thus oriented toward meeting this goal. The management problems include how much habitat to create, whether the Warblers are area sensitive, and whether isolation of habitat patches reduces their likelihood of occupancy. The entire population of Kirtland's Warblers has been censused each spring since 1980. Detailed studies of birth and survival rates have also been conducted. Results from this work have been used to construct a series of viability models varying from simple birth and survival models, with no habitat- or age- based cohorts, to a spatially explicit, individual-based model with a complex movements simulator and age- and habitat-specific birth and survival rates. We will report on the utility of these models for answering the questions above about Kirtland's Warbler recovery and will discuss the applicability of these findings to other species and population viability analyses.

Viability Analysis of a Restored Population of the Federal Threatened Pitcher's Thistle (Cirsium pitcheri) in Illinois. TIMOTHY BELL*, Dept. of Biol. Sci., Chicago State Univ., Chicago, IL, MARLIN BOWLES, Morton Arboretum, Lisle, IL, JENNY MCBRIDE, Morton Arboretum, Lisle, IL, and KAYRI HAVENS, Chicago Botanic Garden, Glencoe, IL.

Pitcher's thistle (Cirsium pitcheri) is a federal threatened monocarpic herbaceous perennial of the western Great Lakes shoreline dune habitats. The plant is self-compatible, with little allozyme variation across its range. Population maintenance of this species depends on cohort replacement and recolonization of successional habitat maintained by shoreline processes. Succession also eliminates populations, while disturbances can either eliminate or create new habitat for populations. These dynamics require metapopulation persistence, in which local populations avoid simultaneous extinction by reacting independently to landscape-scale disturbances and colonizing newly formed habitats. Our population restoration in former Illinois habitat along Lake Michigan comprises Wisconsin, Indiana and Michigan seed sources. Because this species is monocarpic, annual translocation of greenhouse propagated plants was used to build up large cohort numbers. Spontaneous seedlings from flowering plants are now replacing these artificial cohorts, and the first flowering of these plant occurred in 1998. Morphological, demographic, and genetic (as shown by RAPDs) differences occur between geographically different seed sources, with Indiana plants having larger cotyledons and greater growth, survivorship and reproduction in the restoration. The restored Illinois population is now in its seventh year with nearly 150 plants, but population growth rate (lambda) is <1. Stage structured demographic analysis projects cohort sizes, number of spontaneous seedlings and, or, translocations, and amount of flowering needed to sustain population viability. For example, the minimum number of translocated seedlings needed to achieve a positive growth rate is twice the number of currently observed spontaneous seedlings.

How Good Are PVA Models: Testing their Predictions with Experimental Data. GARY E. BELOVSKY, Ecology Center & Dept. of Fisheries & Wildlife, Utah State Univ., Logan, UT.

There are a number of demographic models that have been developed for Population Viability Analysis (PVA). These models have been accepted and applied by conservation biologists although they have not been empirically tested. Testing PVA with the threatened/endangered species for which these models are designed is logistically impossible, and even common species cannot provide the necessary replicated field populations for testing stochastic models. Therefore, I developed an experimental system of populations differing in mean carrying capacity, environmental variability (random variation in carrying capacity) and initial population size that could be sufficiently replicated to test the extinction predictions of stochastic population models used in PVA. The experimental system employed brine shrimp (Artemia franciscana) populations in mesocosms maintained in environmental chambers where a measured food source was added regularly. Standard PVA models did very poorly in predicting observed mean population persistence and its variance. In general, the PVA models: 1) underestimated mean persistence time, and 2) underestimated the influence of carrying capacity and environmental variability. This means that most PVA analyses may be overly optimistic about population persistence. The implications are reviewed for several real conservation cases. Finally, the experimental results for brine shrimp are extended to examine the ability of brine shrimp to withstand current levels of commercial harvesting in The Great Salt Lake.

Effects of Changes in the Demographic Influences of Ocean Conditions on Viability of Pacific Salmon. LOUIS W. BOTSFORD, Dept. of Wildlife, Fish, and Conservation Biology, Univ. California, Davis, CA.

One of the means of improving the projections of population viability analyses is through a better understanding of the effects of demographic uncertainty and environmental variability on extinction probability. Pacific salmon are demographically unique for several reasons and are of great interest due to recent declines and listings. An important, outstanding uncertainty that limits our ability to determine the causes of declines is the influence of ocean conditions on Pacific salmon. Specifically there appears to have been a reduction in ocean survival in the mid-1970s that may have contributed to recent declines. There is also evidence that the age distribution and temporal nature of environmental variability changed. Here we examine how such changes in environmental variability influence probability of extinction using coho salmon and Snake River chinook salmon as examples. Specifically, we explain how various combinations of environmental effects at ocean entry and just prior to the spawning run influence viability of populations with various spawning age structures. The results are useful in planning research to identify mechanisms and in assessing relative jeopardy of various Pacific salmon stocks.

Reconciling the Small Population and Declining Population Paradigms. MARK S. BOYCE, College of Natural Resources, Univ. Wisconsin, Stevens Point, WI.

In a pointed essay in the Journal of Animal Ecology shortly before his death, Graeme Caughley (1994) suggested that conservation biology was plagued with 2 tracks: (1) stochastic models for small populations that have few direct conservation applications, and (2) empirical investigations of declining populations that tend to be case-specific yielding few generalizations. In the 5 years since Caughley's essay, several PVAs have demonstrated that conservation biologists are capable of merging theory and practice. Habitat-based PVAs for California gnatcatchers, grizzly bears, and spotted owls have combined GIS-based spatial analysis with stochastic population models to develop useful links between habitats and extinction risk. Despite progress, many of Caughley's concerns still plague conservation biology, and PVAs in particular. As Caughley pointed out, genetic models are still insufficiently developed and verified to form more than a general basis for conservation management. Theoretical population models are poorly tied to applications because we face major hurdles in developing statistical procedures for fitting models of sufficient complexity to capture the relevant ecology. The very nature of threatened and endangered species means that we seldom have sufficient sample sizes or replication to reliably estimate parameters for complex models. Simple models present fewer problems of statistical estimation, but may have behaviors that are so unrealistic as to be useless for projections into the future. Although adaptive management offered promise for linking science and conservation practice, institutional barriers have proven to be serious impediments. Based on a conviction that science can enhance conservation, we must give urgent priority to Caughley's charge that conservation biologists need to reconcile ecological theory and conservation practice.

Population Viability Analysis for the Marsupial Micoureus demerarae in Small Atlantic Forest Fragments in Southeastern Brazil. DANIEL BRITO*, Depto. Ecologia, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil, and FERNANDO FERNANDEZ Depto. Ecologia, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil.

The fragmentation of the Brazilian Atlantic Forest has left many populations of forest-dwelling marsupials isolated in small forest remnants. Long-term viability of such populations is an open question. A PVA was carried out for populations of the arboreal didelphid Micoureus demerarae in two small (7.0 and 8.8 ha) fragments at Poço das Antas Biological Reserve. Analysis was based on field data obtained through a detailed demographic study carried out at the fragments since 1995. Both populations are very small (averages below 20 individuals) and are connected by dispersing individuals, therefore forming a metapopulation. Frequency of a catastrophe (fire) was estimated from the Reserve's historical records. A thousand simulations were carried out using Vortex. Results so far indicate that both populations taken separately, as well as their metapopulation, have extinction probabilities below 5% for twenty years, but rising to about 96% in a hundred years. The analysis will be improved by incorporating evaluations of genetic variation which are being undertaken at the same fragments. Our final aim is to carry out sensitivity analysis, in order to help understanding what factors are most critical for the persistence of isolated populations of Micoureus demerarae (Funding: Fundação O Boticário de Proteção à Natureza, MacArthur Foundation, FAPERJ, PROBIO).

Evaluating the Predictions of Population Viability Analysis (PVA). BARRY W. BROOK*, Key Centre for Biodiversity and Bioresources, Macquarie Univ., NSW, Australia, RICHARD FRANKHAM, Key Centre for Biodiversity and Bioresources, Macquarie Univ., NSW, Australia, and MARK A. BURGMAN, School of Botany, Univ. Melbourne, Parkville, Victoria, Australia.

Population viability analysis is widely used in conservation biology to predict extinction probabilities and compare management strategies for endangered species. However, it is unclear if different PVA packages produce similar predictions, and whether the predictive models realistically describe the behavior of wildlife populations. Six commonly applied PVA packages, GAPPS, INMAT, RAMAS (Age, Stage and Metapop) and VORTEX, were compared for a range of life-history types. The major findings were: (1) when complex processes were included in the models, large differences were found between some packages, and even versions of the same package. The pattern of similarities and differences altered depending on the species examined. (2) When completely standardized, a consistent difference was revealed between the predicted extinction probabilities of the matrix-based packages (INMAT and RAMAS) compared to those that were individual-based (GAPPS and VORTEX), due largely to a subtle difference in the way demographic stochasticity is modeled. Clearly caution must exercised when interpreting the outcomes of any one PVA package, as their predictions not necessarily concordant. (3) A set of retrospective tests using the historical data of over twenty long-term population studies was conducted, to test the predictive reliability of PVA. It was found that the quasi- extinction and -recovery probabilities predicted by the PVA packages did not realistically reflect the actual population fluctuations. The models also failed to accurately predict future population abundance. (4) A validation and refinement process for PVA was developed, which helped to improve the projections. Current PVA packages cannot be relied upon to produce accurate quantitative predictions.

Demographic Models and Population Viability Analysis: The Surface Has Only Been Scratched. HAL CASWELL, Biology Dept., Woods Hole Oceanographic Institution, Woods Hole, MA.

Like the rumors of Mark Twain's death, recent claims about the limitation of demographic models (based on matrices or other mathematical frameworks) in PVA are greatly exaggerated. Data on endangered species are often limited and uncertain. Both populations and the environments in which they live are complex systems. Neither of these facts suggests throwing away the most powerful tools available for exploring the consequences of those data and the interaction of the environment and the population. I will show how simple matrix population models can be used to quantify the importance of environmental variation, the consequences of demographic as well as environmental stochasticity, and the importance of uncertainty in data. I will briefly consider the concept of "validation", and argue that it is widely misunderstood in the context of demography.

Effects of Individual Heterogeneity in Estimating the Persistence of Small Populations. MARY M. CONNER*, Dept. of Fishery and Wildlife Biology, Colorado State Univ., Fort Collins, CO, and GARY C. WHITE, Dept. of Fishery and Wildlife Biology, Colorado State Univ., Fort Collins, CO.

Demographic, temporal, spatial, and individual heterogeneity are important factors affecting the probability of persistence of small populations. Demographic, spatial, and temporal stochasticity have been incorporated into some Population Viability Analyses (PVA), but individual heterogeneity has not been taken into account. We developed a mathematical simulation model that included baseline demographic stochasticity with varying amounts of temporal and individual stochasticity for different population sizes. As temporal variation increased, the probability of population persistence decreased no matter what the initial population size. However, individual heterogeneity had the opposite effect of temporal heterogeneity. We found that increases in individual variation lead to an increase in the probability of population persistence. Although the increase in persistence in our model was due to random differences in individual survival and mortality probabilities, natural selection provides a mechanistic explanation for this increase of population persistence. Genotypes persisting through a decline may be those that survive better under the conditions causing the decline. Individuals that survive and reproduce in the face of adverse conditions may extend the probability that a small population persists. If variations in life-history traits, such as birth and death parameters, are considered to be affected only by demographic and temporal factors in PVA models, then probabilities of population persistence may be biased low. From a management standpoint, this bias leads to conservative estimates of population persistence in population viability models.

PVA for Plants: Understanding the Demographic Consequences of Seed Dormancy and Disturbance Events in the Face of Limited Data. DANIEL F. DOAK*, Dept. of Environmental Studies, Univ. California, Santa Cruz, CA, DIANE THOMSON, Dept. of Environmental Studies, Univ. California, Santa Cruz, CA, ERIK S. JULES. Dept. of Environmental Studies, Univ. California, Santa Cruz, CA.

Most PVAs rest upon some form of demographic modeling. However, the life histories of many rare plants share two features that greatly complicate the estimation and use of demographic data for viability assessment: seed banks and strong responses to rare and unpredictable disturbance events. These traits are vexing because they are governed by demographic rates that can be extremely difficult to quantify, especially over the short time periods typically available for conservation planning. Furthermore, in many cases these features can mean that census and demographic information for adult populations of plants may yield virtually no information about population viability. Using data from two rare California forbs, in addition to a variety of simulated data sets, we explore the problems of imperfect data and plant life histories for conservation planning. In particular, we emphasize how poor or misleading data on seed banks and disturbance responses can bias estimates of extinction times and population growth rates for plants, as well as for other species. Finally, we use data on broad life history patterns in plants, in combination with simulation models, to evaluate practical techniques for developing PVAs of difficult species with limited data.

Model-based Conservation Decision Making under Ecological Uncertainty. MARTIN DRECHSLER, Dept. of Ecological Modelling, Environmental Research Centre, Leipzig, Germany.

A method of decision making is presented that can be used to compare alternative management actions under ecological uncertainty and to identify which one is likely to have the strongest effect on population viability. The method combines decision analysis with population modeling and uses both information about the population patterns and the population processes. The process knowledge is used to construct the population model and determine plausible ranges for its parameters. The values of these parameters are likely to have an impact on the rank order of the most effective management actions, and unless their ranges are small, there is uncertainty in the management rank order. This uncertainty is encompassed by considering a number of different population parameter combinations, called key scenarios. For each of them a sensitivity analysis is performed and a management rank order determined. In the following decision analysis each key scenario contributes with a certain weight which reflects its biological plausibility. To determine the weight of a particular key scenario, the population dynamics are simulated and the generated patterns are compared with those observed in the real population. The higher the similarity between the two patterns the higher the weight assigned to the key scenario. The decision analysis finally synthesizes the results of sensitivity and pattern analyses and generates a single rank order of the most promising management actions. The method is demonstrated on a case study of the endangered Orange-bellied Parrot Neophema chrysogaster (Australia).

A Population Viability Assessment Model for Captive Breeding Programs. JOANNE M. EARNHARDT, Lincoln Park Zoo, and Dept. of Ecology and Evolution, Univ. Illinois, Chicago, IL

The interaction between wild and captive populations increases as threats to wild populations increase. Captive and wild populations are inexorably linked because captive populations rely on founders either for their colonization or for maintenance of genetic health, and, increasingly, endangered wild populations rely on captive specimens to augment declining wild populations. Because captive and wild populations are tightly bound, the long-term persistence of captive populations is a critical issue for conservation. Management of both types of populations uses computer modeling tools to direct genetic and demographic decisions; however, those tools are not universally applied across the wild-captive conceptual barrier. Methodologies exist for assessing viability of wild populations in our varying environment, but no comprehensive models have been developed for assessment of long-term viability of captive populations. In the development of a Population Viability Assessment model for captive populations, I compared demographic, genetic, and environmental factors under stochastic scenarios in the extinction assessment for both types of populations. Conceptual models for captive and wild populations were diagramed for comparison, the variation in the magnitude of the variables between captive and wild populations was documented, and a preliminary design for captive population viability analysis was developed. Captive populations present unique modeling issues because of the impact of "ceiling"-type carrying capacity and because of genetic and demographic strategies imposed by managers of captive breeding populations. Yet ultimately, captive population viability, similar to wild population viability, will be based on population size, composition of these populations, and on management strategies.

What Do Rare Species Have in Common? REGIS FERRIERE*, Lab. Ecologie, Ecole Normale Supérieure, Paris, France, and Dept. of Ecology and Evolutionary Biology, Univ. Arizona, Tucson, AZ, and OLIVIER RENAULT, Lab. Ecologie, Univ. Paris 6, France.

Rare species have been attracting much attention in recent years, due to an acute concern for their preservation. Since designing conservation plans requires a detailed knowledge of the target species' life history a major question has arisen: do rare species differ in biologically significant traits from common ones? So far insights into this issue have been scarce, disparate and ambiguous. Beyond the fact that reliable data on the biologies of very rare species are exceedingly difficult to collect, the diversity of ways in which studies have been performed (which definition of rarity is applied, how traits are measured, whether environmental differences between populations and phylogenetic effects are accounted for) have greatly confused the analysis of rare-common differences. Perhaps most importantly, no robust theoretical predictions were available to guide empirical work, due to the difficulty of handling models of small populations in which stochastic effects have paramount importance. Here we make use of recent advances in the mathematical theory of multitype branching processes to derive general predictions on life-history traits expected to be associated with rarity. We use the novel framework of adaptive dynamics theory to examine under which conditions rarity may develop as a by-product of natural selection favoring such trait combinations. We confront these predictions with data obtained from the long-term study (14 yrs) of a community of Mexican mountain rattlesnakes (Crotalus willardi willardi, C. lepidus klauberi and C. pricei pricei) known for having the most restricted geographic range among snakes in the U. S.

Population Viability Analysis with Limited Data. JOHN R. FIEBERG*, Dept. of Biomath., North Carolina State Univ., Raleigh, NC, and STEVE P. ELLNER, Dept. of Biomath., North Carolina State Univ., Raleigh, NC.

Population viability analyses are often based on matrix models, made stochastic to account for demographic, environmental, and catastrophic variation in vital rates. One strategy commonly employed is to simulate population projections by randomly choosing among sampled rates. Alternatively, one can use the sampled data to parameterize a probability distribution from which rates can be drawn each year to project the population. Survival and fecundity rates in many natural populations are highly variable from year to year, and therefore, vital rates may have to be sampled for many years before one can accurately characterize the distribution of vital rates (over time). We examine the reliability of conclusions obtained from the above methods given the paucity of data typically available in real life problems. Our results indicate that even a moderate amount of variation in year to year transition rates can seriously undermine the reliability of conclusions drawn.

The Importance of Parameter and Variance Component Estimation in Population Viability Analysis. ALAN B. FRANKLIN*, Colorado Coop. Fish & Wildlife Research Unit, Colorado State Univ., Fort Collins, CO and TANYA M. SHENK, Colorado Div. of Wildlife, Fort Collins, CO.

The results of any population viability analysis (PVA) are predicated on basic parameter estimates, such as abundance or survival, and their associated variance estimates that are used to construct the PVA model. We investigated two problems in parameter estimation that affect the outcomes of PVA: 1) confusing sampling and process variance associated with parameter estimates, and 2) using invalid estimates of parameters. Associated with any parameter estimate are two types of variance: sampling variance (attributable to estimating a parameter from sample data) and process variation (the variation in a given parameter). In the first problem, either sampling variance or total variation (sampling combined with process variation) is often used in lieu of process variance. While the former tends toward slower extinction times when parameters are precisely estimated, the latter tends toward quicker extinction times. In the second problem, invalid parameter estimates can result from, for example, ignoring detectability by using simple counts of animals to estimate abundance and survival. Besides being biased, invalid parameter estimates can also dramatically affect estimates of process variance. The use of simple counts result in estimates that are confounded with detectability. If the true population varies little but detectability exhibits large variation, then parameter estimates of, say, survival may exhibit large temporal process variation solely due to changes in detectability. We illustrate these points with computer simulations, and case studies from the northern spotted owl and Preble's meadow jumping mouse using popular PVA software packages.

Use of Population Viability Analysis in Endangered Species Act Listing and Recovery Decisions for Marine Mammals. LEAH R. GERBER*, Washington Cooperative Fish and Wildlife Research Unit, School of Fisheries, Univ. Washington, Seattle, WA, and DOUGLAS P. DEMASTER, National Marine Mammal Laboratory, Alaska Fisheries Science Center, Seattle, WA.

The Endangered Species Act mandates that Recovery Plans include specific criteria to determine when a species should be removed from the List of Endangered and Threatened Wildlife. In order to meet this mandate, we develop a new approach to determining classification criteria for large whales. The key idea is that endangerment depends on two critical aspects of a population: population size and trends in population size due to intrinsic variability in population growth rates. The way to combine these features is to identify a population size and range of population growth rates (where l denotes the annual multiplicative rate of change of a population) above which there is a negligible probability of extinction. To do so: 1) information on the current population size and its variance are specified, 2) available information on vital rates or changes in abundance over time is used to generate a probability distribution for the population’s l, 3) the lower 5th %-ile value for l (denoted as l _(.05)) is obtained from the frequency distribution of l’s, and 4a) if l _(.05) is less than 1.0, a backwards population trajectory starting at 500 individuals for a period of 10 years is performed and the resulting population size is designated as the threshold for endangered, or 4b) if l _(.05) is greater than or equal to 1.0, the threshold for endangered is set at 500 animals. A similar approach can be used to determine the threshold for threatened under the ESA. We applied this approach to North Pacific humpback whales, with uncertainty in l’s resulting from sampling error and environmental fluctuations. We used Monte Carlo simulations to produce a frequency distribution of l’s for North Pacific humpbacks under three different scenarios. Using the 5th %-ile l, it was determined for the central and eastern populations of North Pacific humpback whales that the best estimates of current abundance were larger than the estimated threshold for endangered, but less than the estimated thresholds for threatened. This analysis leads to a recommendation to downlist the central stock of humpback whales to a status of threatened, while maintaining the eastern and western stock as endangered.

PVA: Questions, Answers, and the Use of PVA. MICHAEL GILPIN, Dept. of Biology, Univ. California, San Diego, CA..

PVA, population viability analysis, is an umbrella concept that covers a general set of models incorporating a probability of extinction at some time in the intermediate future (10-1000 years). There are three general situations that arise repeatedly in conservation decision making: populations that have declined to very low numbers, populations that have been fragmented into isolated patches, and populations that have been harvested or otherwise perturbed. Different types of PVA models must be used in these different cases. The simplest approach of PVA is to base the extinction probability on a projection of current trends. Often required, however, are more mechanistically based PVAs that allow modification of ecological and landscape parameters, so that the impact on extinction probability of a proposed action can be evaluated. These actions may involve exploitation of resources or development that threaten or jeopardize the population, or they may be conservation programs to recover the population or to establish conservation reserves.

Asymptotic Quasi-stationarity of Markovian Models of PVA. FREDERIC GOSSELIN*, Centre d'Ecologie Fonctionnelle et Evolutive, CNRS, Montpellier, France, and JEAN-DOMINIQUE LEBRETON, Centre d'Ecologie Fonctionnelle et Evolutive, CNRS, Montpellier, France.

Contrary to deterministic matrix population models, PVA models lack a theoretical framework to refer to. Among other things, this implies in different publications the periodical "re-discovery" of theoretical results, the varying estimation of similar quantities, and some difficulty in comparing models' results. We here propose asymptotic mathematical results that are sufficiently general to be referred to in most individual-explicit PVAs and population-explicit metapopulation PVAs. These results imply the quasi-stationarity of the model: i.e. the model ultimately almost surely goes extinct, but, if we consider only the non-extinct replicates at any given time, it asymptotically stabilizes in a probability distribution that does not depend on the initial state of the model. This particularly implies that the population has an asymptotic survival rate, as if it were a unique individual, that exhibits the asymptotic interplay between persistence and extinction. Furthermore, different notions of quasi-stability are obtained whether we consider a unique trajectory or a set of many trajectories. Mathematically, these results somewhat resemble those known for deterministic matrix models. They are however a little more difficult to apprehend, but biologically more general, since, among other things, they can apply to models including both demographic and environmental stochasticity, density-dependence, a finite number of different kinds of individuals. We finally discuss the way these results could be usefully applied in PVAs and other individual or population-based models.

A Unifying Approach to Analyzing Persistence of Model Populations. VOLKER GRIMM*, Dept. Ecological Modelling, Centre for Environmental Research, Leipzig, Germany, and CHRISTIAN WISSEL, Dept. Ecological Modelling, Centre for Environmental Research, Leipzig, Germany

Analyzing persistence of small populations has become a key issue in ecology and conservation during the last decade. Unfortunately, a conceptual unification of this field has not yet been developed. Analytical approaches are ponderous and restricted to very simple models. On the other hand, individual-based simulation models allow for more realism but are often not fully explored because of a lack of tools based on theoretical insights. We present a unifying approach which aims at bridging the gap between analytical and simulation approaches. Our approach is based on the notion of "established state" which means a population dynamic characterized by a quasi-stationary distribution of population sizes and by the fact the extinction risk per short time interval is constant. From this notion, we derive the simple but powerful "ln(1-P₀)-plot" with P₀(t) being the probability of extinction by time t. This plot allows us to determine two quantities which are decisive for assessing persistence: T_m,, the mean time to extinction, and P_e, the probability that a model population which is in a certain state will reach the established population dynamic. With T_m, risk of extinction by time t may be calculated for different time frames of interest, and with P_e a threshold population size can be determined below which a population will go extinct quickly. We demonstrate the potential of our unifying approach by presenting analyses of several models addressing particular problems and species.

Pedigree Models for Managing Small Populations in the Wild. SUSAN M. HAIG, USGS Forest and Rangeland Ecosystem Science Center, Corvallis, OR.

Pedigree analysis is perhaps one of the least known, most poorly understood, but useful genetic models for understanding processes in small populations. Pedigree analysis was developed for assessing management strategies in captive populations, but is helpful in understanding processes in natural populations as well. It is a way of avoiding the adverse effects of inbreeding depression by identification of specific individuals contribution to loss of genetic variability in a population. It can be used as a descriptive tool or as a means of testing hypotheses regarding population structure prior to reintroduction or translocation. Pedigree analysis is based on the Gene Drop model, a Monte Carlo simulation where following construction of a simple data set, 10,000 iterations of the model represents sampling an individual's entire genome. Results are summarized via description of: founder contribution, the number of unique alleles surviving, heterozygosity, founder genome equivalents, mean kinship and inbreeding coefficients. The most positive aspects of pedigree analysis are that the data sets required are easy to construct, the programs are easy to run, and because the model takes into account the exact structure of the pedigree, it is quite realistic. The greatest obstacle is the need for a deep pedigree derived from marked individuals. Gene Drop analyses could be considered intermediate analyses between using molecular markers to identify individuals prior to inclusion in a Gene Drop and subsequent use of PVA to further evaluate consequences of introducing individuals identified in the Gene Drop. Examples from our work on Red-cockaded Woodpeckers will illustrate this point.

Modeling Habitat Restoration for San Joaquin Kit Fox: Effects of Patch Size, Immigration, and Stochastic Prey Abundance on Population Persistence. ROBERT G. HAIGHT*, USDA Forest Service, Pacific Southwest Research Station, Albany, CA, PATRICK A. KELLY, Endangered Species Recovery Program, Fresno, CA. KATHERINE RALLS, National Zoological Park, Smithsonian Institution, Washington, DC.

During the last 100 years, the San Joaquin Valley in central California experienced great loss of biodiversity as millions of acres of natural land were converted to cropland and cities. The recovery plan for upland communities of the San Joaquin Valley is largely based on the habitat needs of the San Joaquin kit fox, a nocturnal species that hunts insects, rodents, and rabbits. Remaining kit fox populations live in a few natural areas in the valley and surrounding foothills. The overarching recovery strategy is the establishment of a network of conservation areas linked by rangeland or restored parcels of marginal agricultural lands. Because funds for farmland retirement and habitat restoration are limited, it is important to identify and prioritize areas for restoration. To help this process, we wanted to know how fox populations would likely grow in more-or-less isolated habitat patches. We were concerned about the efficacy of restored areas because they will likely be small (less than 300 square kilometers) and subject to low levels of immigration because of dispersal barriers in surrounding cropland. Furthermore, because of high year-to-year variation in rainfall, prey abundance will likely fluctuate over time. We addressed our concern by developing a simple meta-population model for a local fox population and predicting the relative effects of patch size, immigration, and stochastic prey abundance on a 50-year population trend. The most important results were the effects of stochastic prey abundance, which caused periodic fluctuations in fox reproduction. With constant prey abundance and reproduction rate, fox populations saturated isolated patches that could support 8-32 fox territories. However, with prey fluctuations based on survey data from the Carrizo Plain (near the southern end of the San Joaquin Valley), consequent fluctuations in reproduction rate caused extinctions across a range of patch sizes. Low levels of immigration (1-2 foxes per year) dampened the effects of prey fluctuations and sustained the populations. The results suggested that fox dispersal between patches will be key to sustaining fox populations in restored valley habitat because of fluctuating prey abundance.

Colonization and Extinction Dynamics and PVA. ILKKA HANSKI, Dept. of Ecology and Systematics, Univ. Helsinki, Finland.

Many species have a fragmented population structure and it is hence appealing to consider applying the classic metapopulation concept in their PVA. By the classic metapopulation concept I refer to situations where there is widespread population turnover and where the metapopulation essentially persists in a stochastic extinction-colonization balance. Examples will be given. n other cases population turnover may not be important yet migration may influence local and metapopulation dynamics and hence also e.g., the risk of extinction. if one is primarily interested in extinction-colonization dynamics, models can be parameterized with data that are not too difficult to obtain if one can assume that the metapopulation from which data are collected occurs at a positive quasi-equilibrium. Once the model has been parameterized, it can be applied to other landscapes to predict e.g., expected time to extinction. The problem however is that many metapopulations of endangered species may not occur at a positive quasi-equilibrium in any landscape. Parameterizing models for such situations is much more difficult. In any case, predictions about ultimate extinction times are hardly of great interest, because substantial environmental changes are likely to occur at a faster time scale. Metapopulation models may be most useful for comparing alternative scenarios of landscape management and especially for predicting the occurrence of currently not critically endangered species. For the latter, population-oriented approaches are probably more helpful.

Is Plant Population Viability Influenced by Metapopulation-level Processes? SUSAN HARRISON, Division of Environmental Studies, Univ. California, Davis, CA.

From the perspective of population and metapopulation viability, plants differ from animals in several important ways. Long-distance dispersal is much harder to observe directly in plants, and it may be generally less important in plants than in animals, since plants may resist local extinction through having seed banks. Pollination provides a second route by which spatial isolation may influence plant population viability, however. A review of plant studies shows almost no evidence for the importance of long-distance seed dispersal in population persistence. There is modest evidence that large-scale spatial isolation influences pollination, reproduction, and persistence in plant populations. Metapopulation aspects of viability may only be worth incorporating into models if they are first demonstrated empirically.

A Population Viability Analysis for the Houston Toad. JEFF S. HATFIELD*, USGS Patuxent Wildlife Research Center, Laurel, MD, ANDREW H. PRICE, Texas Parks and Wildlife Dept., Austin, TX, and LISA O'DONNELL, U.S. Fish and Wildlife Service, Austin, TX.

The endangered Houston toad (Bufo houstonensis) is endemic to central and southeast Texas, occurring in nine isolated populations along two separate geologic bands which support deep, friable sandy soils. This species is thought to be declining due to habitat loss, aggravated by catastrophic events such as droughts, and possibly hybridization with other species. Currently, this toad's habitat is protected in only one area, Bastrop State Park (BSP). We analyzed capture-recapture data collected there during 1990-97 and conducted a population viability analysis with RAMAS Metapop software using an age-structured model. Single population models suggest that the relatively low adult female survival rate of 16% per year, combined with low rates of juvenile survival, along with catastrophic events, may be responsible for the decline of this species. Metapopulation models suggest that a relatively low probability of extinction (over 100 years) can be achieved if several populations, each of the size maintained at BSP, are protected and connected by dispersal of individuals (natural dispersal if corridors are maintained, or human-assisted dispersal if corridors are not an option). Increasing the survival rate of tadpoles by controlling predators in the breeding ponds, preventing ponds from drying out before tadpoles have metamorphosed, and possibly increasing the number of ponds, may also help this species recover.

Application of Molecular Genetics to Managing Endangered Species. PHILIP W. HEDRICK, Dept. of Biology, Arizona State Univ. Tempe, AZ.

Molecular genetics can be used to answer questions in conservation that other information may not be able to explicitly address. For example, highly variable genetic markers have been used to determine what are and what are not evolutionarily significant units and to identify populations that have had reduced population size. Examples of both applications will be discussed. However, application of highly variable markers must be used cautiously. In particular, measures of differentiation may be greatly affected by the amount of genetic variation, genetic distance may be greatly influenced by only a few generations of genetic drift, and statistical significance and biological significance may be uncoupled.

Taking a Broader View: Elasticity Analysis of Species Life Histories to Prioritize Research and Management Efforts. SELINA S. HEPPELL, U.S. Environmental Protection Agency, Western Ecology Div., Corvallis, OR.

For most declining populations we do not have detailed demographic or dispersal data to parameterize spatially explicit, stochastic models. While there is heuristic value to running these models with a range of parameters, the number of potential combinations may be limitless for poorly known species. In these situations, we must rely on simpler models (and hence ask simpler questions) and find ways to group species to avoid the need for complete population-specific demographic information. Elasticity analysis of deterministic matrix models gives the proportional contribution of each vital rate (survival, growth and reproduction) to a population's intrinsic rate of increase. While the assumptions of the model restrict it to qualitative predictions about population responses to perturbation, elasticity analysis can be useful for: 1) identifying management options with high or low probability of success, 2) grouping populations or species according to life history traits that are likely to affect their responses, and 3) providing initial information about the sensitivity of different vital rates to guide more complex models or simulations. In a comparative study of 50 mammal life tables, I found that the elasticity of population growth to changes in fertility, juvenile survival and adult survival follow predictions of life history theory and can be estimated from very limited demographic information. Currently, I am using elasticity analysis to categorize amphibian populations according to their likely response to landscape change. This approach, while unable to predict population viability in the strict sense of extinction probability, utilizes readily available information and provides a first step for prioritizing research and management efforts.

Population Extinction: Genetics, Dynamics and Ecology. KEVIN HIGGINS*, Dept. of Biology, Univ. Oregon, Eugene, OR, and MICHAEL LYNCH, Dept. of Biology, Univ. Oregon, Eugene, OR.

New theoretical results on mechanisms of population extinction have focused on synergistic interactions between population genetics and population dynamics. A mechanistic understanding of extinction in terms of the combined impact of population ecology, population genetics and population dynamics is virtually unexplored. Life history and spatial structure have the potential to markedly influence population dynamics and modify the conclusions of earlier work based on a single panmictic population with simple life history. Within this context 1) we investigate how the accumulation of mildly deleterious alleles is affected by barriers to gene flow; and 2) How the interaction between population dynamics and the accumulation of mildly deleterious alleles sets the probability of metapopulation extinction.

A Bayesian Approach to Including Uncertainty in a Hooker's Sea Lion PVA. RAY HILBORN* Sch. of Fisheries, Univ. Washington, Seattle, WA, MIGUEL PASCUAL, Dept of Zoology, Univ. Washington, Seattle, WA, LEAH GERBER, Washington Cooperative Fish and Wildlife Research Unit, Sch.. of Forestry, Univ. Washington, Seattle, WA, FRANCES GULLAND, The Marine Mammal Center, GGNRA, Sausalito, CA, and MARK MAUNDER, Sch. of Fisheries, Univ. Washington, Seattle, WA.

A Bayesian statistical framework is used to calculate the extinction risk to Hooker's sea lion (Phocarctos hookeri) under different levels of incidental mortality in a squid fishery. The analysis explicitly includes (1) meta-population structure with dispersal between existing and potential breeding sites, (2) random catastrophic events that might strike one or more of the breeding sites, (3) depensatory effects causing reduced population growth rates at small population sites, (4) calculation of extinction risk rather than arbitrary pseudo-extinction thresholds, (5) integrated Bayesian analysis of the existing data for Hooker's sea lion, and (6) an analysis of the frequency of catastrophic events and intensity of depensation from similar species. We considered different models of dispersal and suggest that the probability of dispersal should increase with increasing density, and that the probability of dispersal between two sites should decrease with distance. We found that the extinction risk in 200 years under current levels of by-catch is 0.3%, without by-catch it is 0.16%, and with an adaptive by-catch policy in which the fishery would be closed if the population size declined below a threshold it is 0.22%.

Population Viability Analysis and Conservation Policy: Lessons Learned/Lessons Spurned. LAURA HOOD, Defenders of Wildlife, Washington, D.C., INGRID LATCHIS, Defenders of Wildlife, Washington, D.C., BILL SNAPE, Defenders of Wildlife, Washington, D.C., and MARK SHAFFER, Defenders of Wildlife, Washington, D.C.

A qualitative survey of the application of population viability analysis (PVA) in the policy and management arena suggests three major lessons that have been learned, but not acted upon. First, the major limitation to the meaningful application of PVA to conservation problems, namely the lack of detailed population data for many taxa, has been know for 15 to 20 years, yet little has been done to address that weakness. If research (National Science Foundation, U.S. Geological Survey/Biological Research Division) and management institutions (U.S. Fish and Wildlife Service, Forest Service, etc.) had launched long-term, in-depth population ecology studies of carefully selected surrogate or exemplar species, this problem would now be well on its way to a solution. Second, the single "rule-of-thumb" to emerge from the PVA community that seems to have endured the scrutiny of time is Soule's observation that, for vertebrate species at least, populations of less than a few thousand individuals are either in trouble, or likely will be so in the near future. Yet the median population goal to pronounce a species "recovered" is 1522. Third, for all the talk about PVA, and all the many places in federal wildlife law where its concepts could be put to use, there is still no statutory or regulatory definition that provides a meaningful benchmark for society to know how "conserved" any particular taxa is or is not.

The Use of Population Viability Analysis for the Endangered Sonoran Pronghorn Recovery Plan and in a Legal Challenge. DENNIS A. HOSACK*, Defenders of Wildlife, Washington, DC, PHILIP S. MILLER, Conservation Breeding Specialists Group, Apple Valley, MN, JOHN J. HERVERT, Arizona Game and Fish Dept, Yuma, AZ, and ROBERT C. LACY, Chicago Zoological Society, Brookfield, IL.

Sonoran pronghorn (Antilocapra americana sonoriensis) is one of five subspecies of pronghorn and was listed as an endangered species in 1967. The 1982 recovery plan proposed a recovery goal of 300 individuals, although no credible scientific explanation of this recovery goal number has been put forth A Population Viability Analysis (PVA) workshop was organized in 1996 to attempt to provide a more scientifically valid goal. In addition, and pursuant to Section 7 of the Endangered Species Act, Defenders of Wildlife has filed suit against 12 federal agencies for failure to take appropriate steps to protect and recover the Sonoran pronghorn. Current best estimates for the US population of Sonoran pronghorn is 130-160 animals. The results of the PVA suggest that, given the best estimates of current demographic rates, the likelihood of Sonoran pronghorn extinction is 1% within the next 25 years, 9% within 50 years, and 23% within the next 100 years. In addition, the analysis suggested that the population is extremely sensitive to fawn mortality, with the likelihood of extinction increasing markedly when fawn mortality exceeds 70%. Although there was considerable debate among participants about many of the variables and suggestions for recovering the species, it was unanimous that maintaining 95% of the current genetic diversity was a top priority. The results of the workshop analysis indicated that although carrying capacities of 300 animals might be as likely to insure simple survival, only at carrying capacities at or above 500 would the long-term genetic goal be likely to be achieved.

Modeling Equilibrium Population Size and Life-stage Structure for Site-dependent Species. W. GRAINGER HUNT*, Predatory Bird Research Group, Univ. California, Santa Cruz, CA, and PETER R. LAW, Predatory Bird Research Group, Univ. California, Santa Cruz, CA.

Spatial partitioning by pairs of territorial birds restricts cohort size per unit area of landscape, setting an upper limit to an equilibrium size range for the total population deriving from any defined area. Floaters maintain the lower limit by filling territorial vacancies. This phenomenon, known as Moffat's equilibrium, provides the basis for modeling age- and stage structure either under the assumption of annual constancy in vital rates or with variation thereof. The floater-to-breeder ratio at equilibrium is a more informative indicator of population health and stability than a growth rate estimate. The adaptive threshold of site serviceability implied by the existence of floaters (of both sexes) underscores the importance of identifying and conserving the elements of high quality breeding areas for threatened or endangered species. To illustrate these points, we present modeling results of an ongoing study of golden eagle population dynamics in west central California.

Individual Variation Can Influence Population Extinction Risk. HENRIETTE I. JAGER*, Oak Ridge National Laboratory, Oak Ridge, TN and WEBB VAN WINKLE, Oak Ridge, TN

We are conducting a population viability analysis for a fish species with extreme variability in its reproductive life history characteristics. White sturgeon (Acipenser transmontanus) females mature between the ages of 13 and 25 years and at intervals of 2 to 10 years thereafter. Adults with access to the ocean make periodic spawning migrations from the estuary to the river environment. We examined the effects of individual variability in life history parameters on viability predictions from an individual-based population model that incorporated both genetic and environmental variation among individuals. We addressed the following questions: (1) How does individual variation in two traits, age at first maturity and rematuration interval, influence the likelihood of persistence?; (2) Does heritable variation in these traits increase the likelihood of persistence?; and (3) Do changes in the simulated selective regime corresponding to conditions before and after dam construction lead to a shift in mean age at maturity?. The likelihood of persistence predicted by simulations with no variation differed from those predicted by simulations with realistic individual variation in age at maturity. This suggests that it may be important to represent realistic individual variability in life history traits in population viability models designed for species that show high variation. When we simulated a genetic basis for age at first maturity, we found that heritable individual variation increased the simulated likelihood of persistence. Our results raise questions about predicting population responses to man-altered environments with highly modified selective regimes.

Development of a PVA to Identify Research and Management Priorities for a Population of West Indian Manatees in Northeastern Costa Rica. IGNACIO JIMÉNEZ, Regional Wildlife Management Program for Mesoamerica and the Caribbean, Universidad Nacional, Heredia, Costa Rica.

Northeastern Costa Rica harbors the main population of the West Indian manatee in the country and the species is endangered in the area. I developed a population viability analysis (PVA) for the West Indian manatee in northeastern Costa Rica using the software VORTEX to: 1) identify management and research priorities in the area, 2) assess the temporal margin left to local managers to conserve the species, and 3) identify sources and sinks in the local metapopulation. The PVA detected hunting and migration rates, and the carrying capacity of the southern subpopulations as the factors most relevant to population survival. It also showed that survival of manatees in Costa Rica depends on conservation activities carried out in neighboring Nicaragua. Southern subpopulations in the study area act as sources while the northernmost subpopulation behaves as a sink for the whole metapopulation. Research activities to conserve manatees should focus on the animals’ movements and a more precise estimate of hunting rates at both sides of the border between Costa Rica and Nicaragua. Management activities highlighted by the PVA as most needed are: 1) enforcement of laws against poaching and illegal gillnetting in local watercourses, 2) habitat conservation focused in the southern area, and 3) development of bi-national conservation activities between Nicaragua and Costa Rica. Measures to conserve the species should be taken before it falls into a critical situation that could be too difficult to reverse.

Using Demographic Models to Assess the Causes of Recovery: New Insights into the Management of the Peregrine Falcon in California. MATT KAUFFMAN*, Dept. of Environmental Studies, Univ. California, Santa Cruz, CA, and WINIFRED FRICK, Dept. of Environmental Studies, Univ. California, Santa Cruz, CA.

The peregrine falcon (Falco peregrinus) has been touted as a success story of endangered species management. In 1970, the peregrine was on the brink of extinction due to pesticide-induced eggshell thinning. DDT was banned in the USA in 1972 and re-introduction efforts followed. In California, peregrines have increased steadily (140 birds at present, up from 5 in 1970) and were recently de-listed. However, no studies have analyzed whether this recovery is due to: 1) intensive re-introductions, or 2) real demographic improvement due to DDT dissipation. The re-introduction effort has been substantial and sustained—754 birds released since 1977—while long-term monitoring shows no significant declines in eggshell thinning. We conducted a modeling analysis designed to investigate the effects of augmentation and/or an ameliorating environment on the last twenty years of peregrine population growth. To do this, we combined individual records of stage-specific introductions with long-term demographic data to construct two contrasting models: 1) population growth with augmentation only and no accounting for DDT dissipation, and 2) population growth with augmentation and an ameliorating environment reflected by yearly improvement in nest success. We then tested the results of these models against actual population projections using Maximum Likelihood techniques. Our results provide the first quantitative assessment of the effectiveness of the peregrine re-introduction effort, as well as demographic evidence useful in determining whether or not the effects of DDT have been significantly reduced. This work has implications for predicting peregrine population growth and for establishing future management goals for the de-listed peregrine.

Population Viability Analysis for Serengeti Cheetahs. MARCELLA J. KELLY*, Dept. of Wildlife, Fish, and Conservation Biology, Univ. California, Davis, CA, and SARAH M. DURANT, Inst. of Zoology, Zoological Society of London, Regent's Park, London.

Previous modeling attempts for Serengeti cheetahs have produced conflicting results that have implicated either juvenile or adult survival as the primary factor influencing population dynamics. In this study, we provide a thorough analysis of cheetah population dynamics in order determine the viability of the species. We parameterize our model with values based on 20 years of field data and compare our model projections to the actual data. First, we briefly examine the sensitivity of deterministic growth rate. However, populations that on average are stable or have positive population growth, are still subject to stochastic decline or extinction. Therefore we use simulation modeling to analyze extinction risk based on our variance estimates from our field data. In addition we examine the sensitivity of extinction risk to changes in the demographic parameters and to changes in the variance surrounding those parameters. Our results confirm that the population growth rate is most sensitive to changes in adult survival, followed by juvenile survival. In addition, extinction risk is also most sensitive to changes in adult survival. However, in the Serengeti, adult cheetahs are well protected from poaching within the park boundaries while juvenile cheetahs continue to suffer high predation, primarily from lions. We therefore used our demographic records to simulate the effect of different numbers of lions on juvenile survival. High lion abundance dramatically increased extinction risk.

Expanding PVA: Integrating Wildlife Population Biology Models with Models of Human Demographics, Economic Activities, Social Systems, and Other Processes that Impact Biodiversity Conservation. ROBERT C. LACY*, Dept. of Conservation Biology, Chicago Zoological Society, Brookfield, IL, and PHILIP S. MILLER, IUCN/SSC Conservation Breeding Specialist Group, Apple Valley, MN.

Population Viability Analysis uses population biology models to assess probability of extinction, threats, and effects of possible actions. Yet the conservation of biodiversity is mostly a matter of addressing processes that revolve around human populations. Unfortunately, experts who model human systems (such as human demography, local economics and resource use, industrial activities, social systems, and political systems) rarely interact with biologists who use PVA to model wildlife populations. We propose that it is possible and necessary to integrate analysis of human systems with PVA. The linkages between human demographic, economic, and social systems and wildlife population biology must be identified. Outputs from models of each system are necessary inputs as driving forces into other systems. The linkages between human systems and wildlife population biology will be by way of direct harvest of the species, reduction of habitat area, reduction of habitat quality, and habitat fragmentation. Expansive heuristic models can help us identify pathways by which human systems exert pressure on wildlife populations via these threats. Detailed quantitative models of the relevant human systems are needed to generate the specific inputs into a PVA model of the wildlife population. Integrating understanding of human and natural systems will require a broader range of expertise, models, data, and perspectives than has been applied in most PVAs to date.

PVA’s for Populations with Movements Constrained by Landscape Pattern: The Case of the Southwestern Willow Flycatcher. ROLAND H. LAMBERSON*, Dept. of Mathematics, Humboldt State Univ., Arcata, CA, and BARRY R. NOON, Dept. of Fishery and Wildlife Biology, Colorado State Univ., Ft. Collins, CO.

To develop a population viability analysis for the southwestern willow flycatcher, we are exploring multiple models, looking for concordance in model output among competing approaches. A common thread linking the different models is an initial focus on habitat—its amount, quality, and spatial distribution. The tight relationship between the distribution of flycatchers and riparian habitat strongly suggests that movement may be constrained to these linear habitat patches. Such patches can be viewed as fixed nodes of a graph with multiple patterns of connectivity reflecting different "allowable" pathways of movement. Varying assumptions about the degree to which riparian habitat constrains movement uniquely defines different metapopulation boundaries. Our results demonstrate how different assumptions about the "correct" patterns of connectivity affect estimates of persistence and optimal reserve design. Finally, we compare inferences from various connectivity models to traditional demographic-based models, and to incidence functions models parameterized by extensive survey data.

Incorporating Stochasticity and Uncertainty into Population Viability Analysis. RUSSELL LANDE*, Dept. of Biology, Univ. Oregon, Eugene, OR, STEINAR ENGEN, Dept. of Mathematics & Statistics, Norwegian Univ. Science & Technology, Trondheim, Norway, and BERNT-ERIK SÆTHER, Dept. of Zoology, Norwegian Univ. Science & Technology, Trondheim, Norway.

Because population viability is generally expressed as a probability of persistence (above a certain level) for a specified time, stochastic demographic and genetic factors should be incorporated along with deterministic factors into PVA. Deterministic factors include the population effects of habitat destruction and fragmentation, introduced species, overexploitation, and pollution. Stochastic demographic factors include demographic and environmental stochasticity, random catastrophes, and local extinction and colonization in metapopulations. Stochastic genetic factors include random genetic drift and fixation of deleterious mutations. All genetic factors influencing population viability are expressed through demographic factors that affect population dynamics. Uncertainty in estimates of demographic parameters often greatly reduces the statistical confidence that a population achieves any given level of viability, and should cause management to be much more conservative.

Metapopulation Dynamics in Oregon Coastal Natural Coho Salmon Oncorhynchus kisutch: A Buffer against Extended periods of Poor Marine Survival. PETER. W. LAWSON*, Northwest Fisheries Science Center, National Marine Fisheries Service, Newport, OR, and THOMAS E. NICKELSON, Oregon Dept. of Fish and Wildlife, Corvallis, OR.

A habitat-based life-cycle model of Oregon coastal natural (OCN) coho salmon Oncorhynchus kisutch was developed using freshwater production dynamics of individual stream reaches estimated from habitat survey data. Within a river basin, populations in reaches were independent except for straying of five percent of returning spawners. In this sense, each reach represented a population within the basin metapopulation. Monte Carlo simulations showed that during periods of poor marine survival populations in reaches with higher-quality habitat persisted, while those in reaches with low-quality habitat went extinct. When marine survival improved, strays from the persistent populations gradually recolonized unpopulated reaches. Repopulation took several generations, suggesting that OCN coho abundance and distribution may adjust to changes in marine survival over periods of five to ten generations. In healthy metapopulations this time lag could provide resilience against prolonged periods of poor marine survival. Distributions of spawners should be considered along with abundance in evaluating the status of stocks.

A Computer-Intensive Method for Inferring Population Persistence Based on Time Series Data. DANNY C. LEE, Sierra Nevada Conservation Framework, Sacramento, CA..

Population viability analyses often rely on time series data. One approach to modeling such data is use simplistic statistical models to estimate demographic parameters. While analytically elegant, such approaches can overlook important biological factors, including age structure and the interplay between demographic and environmental stochasticity. An alternative approach is to build complex life-history models. Such models simulate population dynamics more mechanistically, but lose the analytical elegance and relations to data of the statistical models. The PopVANC model (Population Viability Analysis based on Nest Counts), uses a hybrid approach to estimating population persistence that relies on the capability of modern computers to readily simulate and analyze millions of synthetic time series. The approach has seven steps: (1) Fit observed time series data using linear regression. (2) Develop a population profile using a fixed life-history structure and observed regression coefficients. (3) Simulate synthetic time series using the predetermined life-history structure and random grid of l (expected replacement rate) and s (environmental noise). (4) Search synthetic data base for observations matching population profile. (5) Build likelihood surface for l and s. (6) Simulate future abundance starting with observed time series and probabilistic draws of l and s. (7) Summarize results as a persistence profile. This approach allows complex life histories, yet provides a rigorous tie to data through the generated likelihood surfaces. The model is illustrated using Pacific salmon redd counts.

Fitting PVA into Adaptive Management. DONALD LUDWIG*, Dept. of Mathematics, Univ. British Columbia, Vancouver, BC, Canada, and CARL WALTERS, Fisheries Centre, Univ. British Columbia, Vancouver, BC, Canada.

Adaptive management initially involves two stages: (1) A modeling stage to identify the main ingredients in understanding the problem and to identify important gaps in existing knowledge, and (2) Design of experiments to try to fill some of these gaps and improve policies. PVA suffers from an apparent requirement for detailed and accurate knowledge in the face of numerous complications due to biological complexity, the complexity of human responses and motivations, and the inherent complexity of the problem of inference. If we underestimate these difficulties in the first stage, we may be led to strategies for the second stage that appear to be cost-effective, but actually impede prudent and practical responses to the problem.

A Reassessment of Florida Panther Population Viability Analysis and Recovery Efforts from Multiple Perspectives. DAVID S. MAEHR*, Dept. of Forestry, Univ. Kentucky, Lexington, KY, ROBERT C. LACY, Chicago Zoological Society, Brookfield, IL, E. DARRELL LAND, Florida Game and Fresh Water Fish Commission, Naples, FL, ORON L. BASS, Everglades National Park, Homestead, FL, and THOMAS S. HOCTOR, Univ. Florida, Gainesville, FL.

We used Vortex (Lacy et al. 1995) to model Florida panther (Puma concolor coryi) population viability from four perspectives. Independent analyses were based on demographic inputs provided by a federal field biologist, a state field biologist, a university conservation biologist, and an NGO population biologist. Despite a lack of full consensus regarding the inputs and results of earlier modeling efforts (Seal et al. 1989, Seal and Lacy 1992), management of this endangered subspecies moved forward first with a plan for captive breeding (which has yet to be implemented) and more recently with genetic restoration. Since 1994, 8 female cougars, introduced from Texas, have produced at least 12 hybrid kittens. Eight may nearly equal the breeding female segment of the population where N_e/N may be as low as 0.26. Because PVAs can use best guesses for model inputs, and because the collection of accurate demographic data from small populations can be a slow, painstaking process, PVA results can easily outpace the collection of data needed to properly drive the model. Panther recovery has been controversial, with the genetic restoration efforts being questioned by analyses which suggest demographic stability of the population may obviate the need for such a radical approach to small population management (Maehr and Caddick 1995). By analyzing the most recent information from different perspectives, we attempt to clarify the status of the panther to increase consensus among managers and to provide guidance as to the most useful recovery actions in the future.

The Demographic and Genetic Contributions of Immigrants in an Insular Song Sparrow Population, AMY B. MARR*, Dept. of Wildlife Ecology, Univ. Wisconsin, Madison, WI, and PETER ARCESE, Dept. of Wildlife Ecology, Univ. Wisconsin, Madison, WI.

Few studies have assessed the true genetic and demographic contributions of immigrants to a population because of practical limits on monitoring effort. This lack of information makes it difficult for conservation biologists to determine objectively how to incorporate the effects of dispersal and inbreeding into Population Viability Analysis models. We used longterm data from a study of an insular song sparrow population to examine the effect of occasional immigration on the genetic composition and fitness of individuals in a small population. Our analysis involved 18 years of data on 313 females for which all breeding attempts were recorded and all offspring were color-banded. Over this period, 16 immigrant females bred and genetic work shows that these individuals introduced novel alleles to subsequent generations. We used our complete pedigree from 1981 to describe the turnover of matrilines and spread of immigrant nuclear genes in the population. Logistic regression was used to determine the consequences of inbreeding, matriline association, and immigrant ancestry for fitness. Our analysis allowed us to estimate directly the genetic and demographic contributions of immigrants, and to suggest how our results might be generalized for inclusion into PVA.

Methods for Testing PVA Models. MICHAEL A. MCCARTHY*, Centre Res. Env. Studies, Aust. Nat. Univ., Canberra, ACT, Australia, HUGH P. POSSINGHAM, Dept. of Env. Sci. and Manag., Univ. Adelaide, Adelaide, SA, Australia, DAVID B. LINDENMAYER, Centre Res. Env. Studies, Aust. Nat. Univ., Canberra, ACT, Australia, and LINDA S. BROOME, NSW Nat. Parks & Wildl. Serv., Queanbeyan, NSW, Australia.

Despite being first developed approximately 20 years ago, the predictions of PVA models have rarely been tested. We present several different methods for testing PVA models, which aim at identifying their weakest aspects. The different methods require different sorts of data, such as presence/absence data in patches and time series data of population size. Both graphical and statistical methods are used. The methods are illustrated with examples of PVA models of mountain pygmy-possums, arboreal marsupials and Australian treecreepers. Three different types of model are analyzed: single population models; patch-occupancy metapopulation models; and metapopulation models that include local dynamics. In most cases, there is at least reasonable congruence between the predictions of the models and the observed data.

Population Viability Analysis of an Endangered Passerine: Computer-modeling Results for a Translocated Metapopulation. ANDREW MCCLUNG*, Dept. of Zoology, Univ. Hawaii at Manoa, Honolulu, HI, MARIE MORIN, Dept. of Zoology, Univ. Hawaii at Manoa, Honolulu, HI, and SHEILA CONANT, Dept. of Zoology, Univ. Hawaii at Manoa, Honolulu, HI.

Population Viability Analysis (PVA) has proven useful in assessing management options for endangered species, especially when the biology of the species is well-known. Most studies to date have focused on continental species and habitat fragmentation, but few have addressed management options for relict species in island ecosystems. The Northwest Hawaiian Islands contain several passerine birds that could be so described, and formerly contained many others since gone extinct. Our research, using the VORTEX PVA modeling program, indicates that one of these birds, the Laysan Finch, may be in the process of losing a group of populations at Pearl & Hermes Reef (PHR). Translocated to PHR in 1967 as insurance against catastrophic events at Laysan, the PHR finches face environmental stochasticity comparable to that affecting their parent population, but patch sizes and effective population sizes at PHR are much smaller. Also, local population extinctions remain problematic due to low inter-islet migration rates and consequent lack of population replenishment. Furthermore, the vegetation composition has changed significantly over the last 20 years, creating a severe limitation on nest sites and shifts in behavior that may be affecting breeding success. Preliminary VORTEX results suggest that inbreeding is much more a problem at PHR than at Laysan, and in concert with earlier genetic studies, indicate that the PHR populations provide neither a sufficient reserve of genetic diversity nor of demographic stability. Whether to respond to this decline by attempting an additional translocation at a more stable island habitat remains to be answered.

Linking the Human Animal with Endangered Species Risk Assessment: Case Studies of an Expanded PVA Process. PHILIP S. MILLER*, IUCN/SSC Conservation Breeding Specialist Group, Apple Valley, MN, FRANCES WESTLEY, IUCN/SSC Conservation Breeding Specialist Group, Apple Valley, MN, and ANN P. BYERS, IUCN/SSC Conservation Breeding Specialist Group, Apple Valley, MN.

The use of demographic and genetic models of the extinction process has become an increasingly valuable component of endangered species management. However, due in large part to great consternation amongst biologists concerning the conceptual foundations and detailed mechanics of these methods for population viability analysis (PVA), we neglect to concern ourselves with perhaps the most important element in applying PVA to real conservation action: the link between human population processes (demographic, economic, and socio-political) and the risks of wildlife population extinction. Just as these human processes are the principal agents of wildlife population endangerment, so too must they be the subjects of rigorous analysis as part of a broader risk assessment. Through the Population and Habitat Viability Assessment workshop process, we are attempting to augment the standard PVA approach by working to expand stakeholder involvement and to incorporate quantitative estimates of human population growth and land use projections into the modeling process. We have implemented this expanded process in four countries representing a diverse array of socio-cultural domains. Two specific case studies are described in some detail: the mountain gorilla (Gorilla gorilla beringei), where parameterizing and simulating the effects of disease outbreak and violent civil unrest around the Virunga volcanoes region formed the bulk of the modeling effort; and the tree kangaroos of Papua New Guinea (Dendrolagus sp.) in which considerable semantic knowledge from local villagers was assembled to evaluate the impacts of hunting on locally fragmented populations. Lessons learned from these initial experiments with an expanded process are discussed.

Sensitivity Analysis to Evaluate the Consequences of Conservation Actions. L. SCOTT MILLS, Wildlife Biology Program, Univ. Montana, Missoula, MT.

Sensitivity analysis is a critical component of PVA, because instead of merely flagging low persistence probability it helps us identify specific activities that are most likely to improve population growth or persistence. The three main methods for conducting sensitivity analysis include: a) elasticities and LTRE approaches; b) life-stage simulation analysis; c) approaches based on PVA analysis. I will address the potential and pitfalls of each of these, emphasizing the problems that arise when "best guess" or mean estimates are used without accounting for variation in birth and death rates. To date, sensitivity analysis has not yet been extended to multiple populations, although levels of connectivity (or dispersal rates) are considered to be key parameters in demographic and genetic analyses of metapopulations. As an example using real field data, I use a simple model based on lesser snow geese to demonstrate how ignoring connectivity, or ignoring a plausible range of variation in vital rates, can lead to misleading management recommendations of the importance of particular conservation actions for metapopulations. The pivotal role of variance in evaluating the importance of within and among-population vital rates illuminates the need for field estimates that separate process variation from sampling variation.

Regional-scale Habitat Modeling When Data are Insufficient for Population Viability Analysis. REED NOSS*, Conservation Biology Institute, Corvallis, OR, CARLOS CARROLL, Conservation Biology Institute, Corvallis, OR, PAUL PAQUET, Conservation Biology Institute, Corvallis, OR, and JAMES STRITTHOLT, Conservation Biology Institute, Corvallis, OR.

Developing conservation plans for wide-ranging species, such as large carnivores, requires information on the distribution of suitable habitat over large regions, from which inferences about population persistence can be made. Typically, in such cases, the demographic information necessary for credible population viability analysis is lacking. Furthermore, existing conceptual models of habitat suitability, based on fine-scale habitat attributes, are not useful for regional-scale modeling. We are developing empirical GIS-based models of habitat quality for 12 species of carnivores in the Rocky Mountains of the U.S. and Canada based on "found" distributional data such as trappings and sightings, as well as regional-scale attributes predicted to be important to long-term persistence. Although dynamic simulation models may offer the strongest mechanistic insights, the poor knowledge base for most of these species is inconsistent with these models' sensitivity to error-propagation. The use of static spatial models, such as the autologistic, allows analysis of habitat requirements across multiple scales in the absence of detailed demographic data. Commonalities among species in their response to habitat attributes facilitates the development of multi-species models and conservation strategies. Our modeling approach is iterative, with initial models offering information for short-term land-use and conservation decisions, as well as helping to identify key knowledge gaps and prioritize future studies. An "adaptive monitoring" approach allows initial empirical models to be validated and revised using new field data.

The Use of PVA to Compare the Relative Effects of Alternative Proposed Hydrologic Scenarios on the Everglades' Cape Sable Seaside Sparrow. PHIL NOTT, Institute for Environmental Modeling, Univ. Tennessee, Knoxville, TN.

SIMSPAR is a spatially-explicit individual based model designed as a management tool for the Cape Sable Seaside Sparrow (Ammodramus maritimus mirabilis) of the Florida Everglades. It represents one of the models within the Across Trophic Level System Simulation (ATLSS) Program of the U.S. Geological Survey, Biological Resources Division, an Everglades ecosystem-wide model. The model includes landscape layers of habitat and topography at 500-m resolution upon which individuals in a small population respond to the changing landscape and encounters with other individuals. The life history and behavioral characteristics of model individuals are based on field observations of the species. Spatial and temporal patterns of breeding success are mainly determined by historical daily water levels affecting the availability of breeding habitat. The model was validated using historical daily water levels as input and comparing the resultant temporal and spatial patterns of breeding with those known from field surveys over a 15-year period. The model is now used to predict the impact of proposed alternative hydrologic scenarios on Cape Sable Seaside Sparrow populations. The processes of mortality, mate choice, and dispersal are expressed stochastically; hence the model is implemented as Monte Carlo simulations. Population viability analyses are applied to the variable results to determine population variability and the risk of explosion or extinction. This allows comparison of the relative merits of one or other management scenarios.

Wolves, Deer, and Logging: Population Viability and Predator-prey Dynamics in a Disturbed Insular Landscape. DAVID K. PERSON*, Alaska Dept. of Fish and Game, Ketchikan, AK, and R. TERRY BOWYER, Institute of Arctic Biology, Univ. Alaska Fairbanks, Fairbanks, AK.

Analysis of population viability typically examines the demography and habitat relations associated with a single species. The population dynamics of obligate predators such as wolves (Canis lupus) are inextricably linked with those of their prey, and viability analyses for these carnivores should address influences on the predator-prey system rather than simply focusing on the predator population. We describe a population viability analysis for wolves that employs a predator-prey model to predict the effects of the revised Tongass National Forest Management Plan on the persistence of the wolf-deer-human system on Prince of Wales Island in southeastern Alaska. We conclude that it is unlikely that the current plan for the Tongass National Forest will result in the extinction of wolves from Prince of Wales Island in the next century. Nevertheless, the wolf population may decline 30-40% over the next 100 years and deer populations will likely decline disproportionately to the decay of carrying capacity. Prior to the initiation of industrial-scale timber harvesting in 1954, carrying capacity was sufficiently high to enable wolf and deer populations to rebound after severe winters. Long periods in which deer numbers are suppressed by predation will likely occur in the future because of the loss of carrying capacity for deer. Declining populations of deer will result in conflicts between hunters and wolves.

Model Choice in Population Viability Analysis. CATHERINE A. PFISTER, Dept. of Ecology and Evolution, Univ. Chicago, Chicago, IL.

All modeling techniques that we use to examine population viability require we decide what level of detail will be included in the model. Should individual data be used or should we use aggregate descriptors for our populations? We often decide based on the quantity and quality of data available for the species we are studying, rather than a knowledge of how the model formulation will affect our predictions. Using a series of analyses that compares the outcome of individual based models (IBMs) with models that aggregate individuals into size classes (matrix projection models), I have found features of natural populations that cause matrix models to deviate from IBMs. In both density dependent and independent scenarios, populations composed of individuals with strongly size-dependent performance and the tendency for individuals to have positive correlations in performance through time were always poorly described by matrix models. While the dynamics of populations reduced to small size were also better described by IBMs, high levels of stochasticity did not favor the use of a particular model. These results provide guidance for how we might better match the characteristics of a species of concern and the model needed to accurately describe it.

Decision Theory and Population Viability Analysis. HUGH POSSINGHAM, Dept. of Applied & Molecular Ecology, Univ. Adelaide, Australia and National Center for Ecological Analysis and Synthesis, Santa Barbara, CA .

The science in our conservation journals has limited application to applied managers because it has not been developed within a decision theory framework. Population Viability Analysis is a classical example of a theoretical tool developed with the limited (if not flawed) purpose of estimating the viability of populations. The real benefits of PVA arise when we use it to explicitly evaluate decisions and/or rank management options. Costs, benefits and trade-offs should be the cornerstones of our thinking. In this paper I will briefly review some applications of PVA in a decision support role. I then explore the potential application of more formal decision-making tools, like stochastic dynamic programming, for finding state-dependent management regimes for threatened species. Relatively simple stochastic population models are all that is needed to make fairly robust decisions. Using this approach we can explore questions in conservation ecology like: Where should habitat be restored - adding to existing patches or making corridors? When should an endangered species be brought in to captivity? How should fire be managed? What forest management strategies minimize extinction probabilities for low cost? Answers to these questions will lead us to a theory of applied conservation, not more theory.

Extinction Probabilities of a Rare Species under Variable Fire Regimes PEDRO F. QUINTANA-ASCENCIO*. Archbold Biological Station, Lake Placid, FL and El Colegio de la Frontera Sur, San Cristóbal de Las Casas, Chiapas, 29200, México and ERIC S. MENDES, El Colegio de la Frontera Sur, San Cristóbal de Las Casas, Chiapas, 29200, México.

We used matrix population models to compare Hypericum cumulicola demography and extinction probability in Florida scrub under different fire scenarios. Demographic data were obtained from five-year census of 15 populations (1994-1998; 3300 plants), field observations, and experiments. Projected finite rates of increase ranged between 0.597 and 3.69. Time-since-fire and finite rates of increase showed a significant inverse relationship. Expanding populations were predicted mostly in recently burned patches. Growth was the most important elasticity component in recently burned patches while soil seed bank survival had the largest elasticity in sites a decade postfire and in long-unburned patches. Simulations of population trajectories after an initial fire indicated that in the absence of migration and additional fires, even large populations may become locally extirpated within 200 years. Fire return intervals less than 20 years and variation in fire return intervals are recommended as appropriate management for this rosemary scrub endemic.

Guidelines for Using PVA for Endangered Species Management. KATHERINE RALLS*, National Zoological Park, Smithsonian Institution, Washington, DC, STEVEN BEISSINGER, Dept. of Environmental Science, Policy, and Management, Univ. of California, Berkeley, CA, JEAN FITTS COCHRANE, Dept. of Ecology, Evolution, and Behavior, Univ. Minnesota, St. Paul, MN, and ANTHONY M. STARFIELD, Dept. of Ecology, Evolution, and Behavior, Univ. Minnesota, St. Paul, MN.

PVA models use a wide variety of approaches to understand past population trends, evaluate likely threats to a population, and project future population trends. However, the formal population models used in PVA are not the only useful approach to recovery planning. PVA is widely used to gain a better understanding of an endangered species and its problems and to estimate the relative risk of alternative management options. There are no consensus guidelines on why, when, or how PVA should be used in endangered species management. We describe PVA and alternative approaches to gaining understanding, estimating risk, and making decisions. It is crucial to define objectives before choosing a specific technique. We provide guidelines to help decision-makers decide whether or not a PVA would be helpful for a given species, choose or develop an appropriate model, conduct the analysis, and interpret the results. We also develop criteria for judging the quality of a completed PVA. Although many PVA's are conducted during brief workshops and never revisited, PVA is more useful as a long-term, iterative process coupled with an adaptive management approach to species recovery.

The Use of Demographic Data and a Spatially Explicit Population Model to Analyze Effects of Habitat Management on Persistence of Northern Spotted Owls on the Olympic Peninsula, WA. MARTIN G. RAPHAEL*, U.S. Forest Service, Pacific Northwest Research Station, Olympia, WA, RICHARD S. HOLTHAUSEN, U.S. Forest Service, Rocky Mountain Research Station, Flagstaff, AZ, and KEVIN McKELVEY, U.S. Forest Service, Rocky Mountain Research Station, Missoula, MT.

We analyzed likely patterns of distribution and persistence of northern spotted owls (Strix occidentalis caurina) on the Olympic Peninsula, WA. We used a spatially explicit population simulation model for northern spotted owls for the analysis and also reviewed current information on demographics and likely owl population numbers on the peninsula. Where there was uncertainty surrounding the values of input parameters, such as the relationship between habitat quality and survival and fecundity rates, we looked at the sensitivity of model results to variation in parameter values. Analysis focused on the effects of Federal habitat under provisions of the Northwest Forest Plan; additional benefits to the owl population of different levels of habitat retention on non-Federal lands; effects of establishing a habitat connection between the Olympic Peninsula and other parts of the owl's range; the likely rate of habitat regrowth in the National Forest and its effect on the owl population; and the likely effect of a worst-case fire. We concluded that the retention of non-Federal habitat would make a biologically significant contribution to the maintenance of the population on the peninsula. We also concluded that a habitat connection across Southwest Washington, based on a design suggested by the Northern Spotted Owl Recovery Team, would have little effect on the status of the owl population on the peninsula. This analysis demonstrated the value of spatially explicit population modeling based on locally-collected demographic information. However, even with a long-term data base and powerful analytical models, significant uncertainty remained concerning future projections, and we had to resort to professional judgment to decide which analytical scenarios were most likely.

Determining Viable Stream Networks for Persistence of the Lahontan Cutthroat Trout. CHRIS RAY*, Dept. of Biology, Univ. Nevada, Reno, NV, MARY PEACOCK, Dept. of Biology, Univ. Nevada, Reno, NV, and JASON DUNHAM, Dept. of Biology, Univ. Nevada, Reno, NV.

Once found in cold streams and lakes throughout the north-central Great Basin, Lahontan cutthroat trout have experienced a 90% range reduction in the last century. Most contemporary populations are isolated in minor tributaries, due to the degradation of lake and mainstem stream habitats. Stream flows in the Great Basin are extremely variable in space and time, and previous analyses suggest that historic populations persisted through frequent redistribution within stream networks. Here we develop models to predict the viability of networked populations, using data from long-term studies of age-structured recruitment. Data were gathered from several isolated populations and from one of the few remaining population systems connected by a functional mainstem. Lacking specific data on dispersal, we adopt two approaches to modeling network dynamics. 1) We incorporate dispersal implicitly by using data from the observed network to determine relationships between network 'state' (average demographic and environmental variables) and the growth of individual populations. The dynamics of arbitrary networks are predicted by linking populations that obey these relationships. 2) We use data from isolated populations to determine relationships between population state and growth. Larger networks are modeled by linking populations that obey these relationships and adding hypothetical levels of dispersal. We compare assumptions and predictions between approaches, and discuss the potential for persistent trout networks in watersheds throughout the Great Basin.

Application of Population Viability Analysis in the Development of the Wisconsin Wolf Management Plan. ROBERT E. ROLLEY*, Wisc. Dept. Nat. Res., Monona, WI, ADRIAN P. WYDEVEN, Wisc. Dept. Nat. Res., Park Falls, WI, and BRUCE E. KOHN, Wisc. Dept. Nat. Res., Rhinelander, WI.

The gray wolf (Canis lupus) was extirpated in Wisconsin in the 1950s but recolonized in the mid 1970s. The Wisconsin Department of Natural Resources (WDNR) listed wolves as state endangered in 1975 and implemented a recovery plan in 1989. By 1997, the wolf population had met the recovery goal of 80 or more wolves for 3 consecutive years. The WDNR is currently proposing to reclassify wolves as a threatened species, and is developing a plan for their management as a threatened and eventually as a delisted species. We analyzed the viability of the Wisconsin gray wolf population using VORTEX software to aid in plan development. Radio-telemetry studies provided estimates of mean reproduction and mortality rates. We evaluated several environmental variability and catastrophic event scenarios because data on these parameters were limited. Virtually all simulations with delisting criteria of 100+ wolves resulted in population persistence when we assumed low to moderate environmental variability and <5% probability of catastrophic event. Estimates of risk of wolves declining below the recovery goal varied little among delisting criteria of 200-500, but were very sensitive to magnitude of environmental variability and the probability of catastrophic events assumed. Our analysis emphasizes the importance of continued monitoring in the development of an adaptive management program for Wisconsin's wolves. The long- term persistence of wolves in Wisconsin depends on finding the proper balance between wolf protection and control to address public concerns about livestock and pet depredations.

Population Viability of a Small, but Rapidly Increasing Arctic Goose Population. MARCUS ROWCLIFFE*, Inst. of Zool., Zoological Soc. London, London, UK, RICHARD PETTIFOR, Inst. of Zool., Zoological Soc. London, London, UK, and JEFFREY M. BLACK, Dept. of Wildlife, Humboldt State Univ., Arcata, CA.

Use of Population Viability Analyses (PVA) for predicting population stability can lead to unrealistic conclusions if models are based on poor-quality demographic data. Because data from small, endangered populations are often meager, it is difficult to assess the usefulness of the PVA approach as a conservation tool. We attempt to validate the PVA approach by utilizing data from an intensively studied small population of Barnacle Geese Branta leucopsis. This population is among the most celebrated wildlife management and conservation success stories in the United Kingdom and Norway It was through proactive wildlife policy and the establishment of well-managed refuges that the population recovered from critically low numbers; i.e. 300 in 1947 to 24,000 in 1997. Based on the 25-year dataset of individually marked birds (i.e. their survival and reproductive success) and other population parameters, we assess the risk of extinction and potential growth within a given time span. We consider a range of range of population responses based on potential management scenarios (e.g. further improving refuge management policy, allowing harvesting etc.) while considering the severe stochastic nature of the arctic breeding grounds. We conclude that the PVA approach is indeed a useful conservation/management tool when appropriately detailed data is available.

Understanding the Importance of Stochastic Processes in PVA: Estimates of Some Essential Parameters. BERNT-ERIK SÆTHER*, Dept. of Zoology, Norwegian Univ. Science and Technology, Trondheim, Norway and STEINAR ENGEN, Dept. of Mathematical Sciences, Norwegian Univ. Science and Technology, Trondheim, Norway.

Fluctuations of small populations are likely to be strongly influenced by different forms of stochasticities. A problem in making reliable predictions from population viability analysis is to obtain estimates of the stochastic parameters in the model. Here we will present formal definitions of demographic and environmental variances. Based on methods derived from these definitions, we will present estimates of these two parameters in three populations of small temperate passerines (great tit, European dipper and song sparrow) where longterm data on both population fluctuations and individual variation in recruitment success were available. We will then calculate the expected time to extinction, applying a diffusion approximation, and examine which demographic factors most likely to affect the probability of extinction in these three populations. We will then discuss the applicability of these results for viability analysis of rare or endangered species where such long-term data are rarely available.

Reintroduced Population Viability Analyses: What Can Be Done? FRANÇOIS SARRAZIN*, Lab. of Ecology, CNRS -UMR 7625, Univ. Pierre & Marie Curie, Paris, France, and STEPHANE LEGENDRE, Lab. of Ecology, CNRS -UMR 7625, Ecole Normale Superieure, Paris, France.

Reintroductions are more and more often used for conservation but they remain too rarely successful. Since the aim of any reintroduction is to restore a self-sustaining population, this is mainly a matter of extinction avoidance. In that context, Population Viability Analyses may be useful in two ways. First they may help to compare a priori reintroduction strategies. However, since the quality of data available for species candidate to reintroduction is often questionable, PVA have to be considered as formal approaches providing qualitative results Missing data can nevertheless be inferred from life cycle and habitat quality scenarios. Second, once the reintroduction is running, PVA may help to assess the success of the project and to plan the management of the reintroduced population. These models have to consider both life history traits and release parameters (number and duration of releases, origin and age class of individuals, habitat quality). They therefore necessitate highly plastic modeling tools. Using the software ULM, we developed two-sex models including demographic stochasticity to assess the consequence of breeding systems on reintroduced populations viability for different release parameters. Furthermore, we compared the relative efficiency of releasing juveniles or adults for a given life cycle. This family of models accounted for possible reduction of survival and fertility of released adults . Such a model was applied to the case of the reintroduction of Griffon vultures Gyps fulvus in Southern France for which accurate monitoring allowed us to estimate demographic rates and consequences of releases.

The Implications of Territorial Clustering on the Demography and Dispersion Patterns of Small Populations of Birds. THOMAS SCOTT, Dept. of Environmental Science, Policy,. and Management, Univ. California, Berkeley, CA..

The San Clemente Loggerhead Shrike remains the one of the rarest North American subspecies of passerine. By any measure, the subspecies persists at the conventional limits of population viability (7 to 18 pairs during the 15 years of our study). Nevertheless, the persistence of this population and its isolation on San Clemente Island (southern California) have provided an opportunity to study long-term demography and dispersion patterns of small populations without the confounding influence of immigration. Many parameters of the population seem driven by the longevity of breeding males and their use of life-long territories rather than individual variance in reproductive success. Female survivorship, emigration, and movement (among territories) appear to be strong contributors to male territorial clustering in shrikes. In general the viability of the population has been most strongly influenced by males longevity and female choice of mate/breeding territory. Nesting pairs were more clustered than would be predicted by habitat quality, with strong implications for meta-population structure and landscape-level dispersion patterns of other passerines. The spatial pattern can either be viewed as a manifestation of declining populations or as an intrinsic nesting pattern magnified by island isolation. In either case, clustering may play a substantive role in the survival or extirpation of small populations of birds.

Estimating the Magnitude of Environmental and Demographic Stochasticity in Survivorship Data for a Nearctic-Neotropical Migrant Songbird. T. SCOTT SILLETT*, Dept. of Biological Sciences, Dartmouth College, Hanover, NH, SCOTT A. MORRISON, Dept. of Biological Sciences, Dartmouth College, Hanover, NH, KATHRYN L. COTTINGHAM, Dept. of Biological Sciences, Dartmouth College, Hanover, NH, RICHARD T. HOLMES, Dept. of Biological Sciences, Dartmouth College, Hanover, NH.

An explicit accounting of stochastic effects on survivorship is critical to sound population viability analysis. Annual variation in survival is typically attributed to environmental factors. However, assuming that all variation in survivorship is a result of environmental stochasticity can lead to overly pessimistic predictions of population viability. We used the method of Kendall (1998: Ecol. App. 8:184) to separate demographic and environmental stochasticity in survivorship of a Nearctic - Neotropical migrant songbird, the black-throated blue warbler (Dendroica caerulescens). Data were gathered from 1986-1998 on two warbler populations: one in the breeding grounds in New Hampshire, USA, and one in the non-breeding quarters in Jamaica, West Indies. The black-throated blue warbler is representative of many migrant passerines in its life history and demography, and our sample sizes are as large as those from most published bird studies. Based on our analyses, all variation in survivorship could be attributed to demographic stochasticity, despite wide variation in annual survivorship (0.3 - 1.0) for both populations. Simulations to explore the role of sample size and variance in survivorship on the statistical results indicate that our power to detect environmental stochasticity, if it existed, was low (<0.3 for both populations). This suggests that very large sample sizes or more extreme variation may be needed to produce meaningful results. We therefore recommend caution in applying Kendall's (1998) method, especially to smaller datasets typically available for species of conservation concern, where policy decisions might not wait for longer-term data collection.

Error Propagation, Population Viability Analysis, and the Natural Boundaries of Knowability. DAVID SMITH*, Dept. of Biology, Univ. Maryland, College Park, MD, JON PAUL RODRIGUEZ, Dept. of Ecology and Evolutionary Biology, Princeton Univ., Princeton, NJ, and STEVE PACALA, Dept. of Ecology and Evolutionary Biology, Princeton Univ., Princeton, NJ.

Population viability analyses (PVA) depend on estimates of basic demographic parameters, but measurement errors are often ignored. We collected published life-tables and used several methods to generate pseudo-data. We investigated the effect of measurement errors on risk assessment by bootstrapping the data and propagating the errors through a stochastic, demographic model. The risk from error propagation was compared to the risk from demographic stochasticity by repeating the stochastic, demographic simulation using the estimates as if they were true values. Error propagation had a larger impact than demographic stochasticity when populations were small and had growth rates close to zero--in other words, those populations where a PVA could be the method of choice for assessing the relative value of alternative conservation strategies. Because studies of small populations must have small sample sizes, there are limits on what can be known about small populations. We identify the bounds to the unknowable region in terms of population size, longevity, growth, and several other aspects of the demography. This study has implications on an important question--should developers or conservationists have the burden of proof in a conflict over a small population?

Desensitizing and Constraining Dispersal in an Individual Based Model. BRADLEY M. STITH*, Dept. of Wildlife Ecology and Conservation, Univ. Florida, Gainesville, FL, JOHN W. FITZPATRICK, Cornell Laboratory of Ornithology, Ithaca, NY; and GLEN E. WOOLFENDEN, Archbold Biological Station, Lake Placid, FL.

We examined the sensitivity of an individual based model for the Florida Scrub-Jay (Aphelocoma coelurescens) within artificial landscapes configured to maximize the influence of dispersal on population persistence. We compared the relative importance of total metapopulation size, local patch size, interpatch distance, and dispersal ability. Local patch size was much more important than the other 3 factors, even though maximum patch size was small, supporting no more than 8 pairs of jays. In contrast to several recent studies, our model was relatively insensitive to substantial differences in dispersal parameter settings, even where dispersal should be maximally important. We attribute this model robustness to design features of the dispersal module and increased model realism. We also describe how different types of field data can be used to place effective constraints on plausible parameter settings. Long term research at Archbold Biological Station provided data that model output could be tested against. Radiotracking provided rough estimates of disperser movement ability and survival. Knowledge of where colonizations should NOT occur placed further constraints on dispersal parameters. We suggest that PVA models are not all overly sensitive to dispersal parameters, and often can be constrained by a variety of field data. Moreover, increasing model realism may further constrain model behavior; error propagation can actually decrease with increasing model complexity.

Genetic Based Effective Population Sizes: Are They? BRADLEY J. SWANSON, Dept. of Biological Sciences, Purdue Univ., W. Lafayette, IN.

One component required for long term population viability is maintenance of genetic variation. The rate of loss of genetic variation is controlled by genetic drift, the strength of which can be estimated by Wright’s effective population size (N_e). N_e can be directly measured with genetic markers such as allozymes and microsatellites. In this paper I examine how N_e varies when estimated from allozyme heterozygosity or microsatellite heterozygosity. N_es were estimated from published reports of heterozygosity using the infinite allele model for allozymes and the stepwise mutation model for microsatellites with mutation rates of 10^-7 and 10^-3 respectively. I found that the average allozyme-based estimate of N_e (151,000 + 26.3 SD) based on 682 populations from 152 mammalian species was significantly greater than the average N_e for microsatellites (550 + 1.14) based on 132 populations from 36 species. In paired species comparisons the ratio of allozyme estimates to microsatellite estimates was greater by 3 orders of magnitude. In an intraspecific comparison of 72 populations of white-footed mice (Peromyscus leucopus) the average allozyme estimate of N_e (912,000 + 2.34), was 200 times greater than the microsatellite estimate of N_e (5750 + 2.4). The average N_e/N ratio for populations with allozyme data and census sizes was 829 (+ 1376). N_es estimated from allozymes seem too high to be accurate. Results were robust to changing both mutation model and mutation rate. Estimates of N_e based on allozymes were also several orders of magnitude greater than those estimated from temporal change in heterozygosity and demographic methods. This work suggests that temporal methods are most appropriate when estimating N_e over the time periods important for population viability.

Use of PVA Models for Ranking Species Vulnerability. BARBARA L. TAYLOR*, Southwest Fisheries Science Center, La Jolla, CA, PAUL WADE, National Marine Mammal Laboratory, Alaska Fisheries Science Center, Seattle, WA, UMA RAMAKRISHNAN, Dept. of Biology, Univ. California, San Diego, CA, and MICHAEL GILPIN, Dept. of Biology, Univ. California, San Diego, CA..

We use simulations to evaluate two types of Population Viability Analysis (PVA) for risk classification of species. We perform an experiment where one author creates multiple data sets at varying levels of risk and submits these data to the other authors for analysis and risk classification. We contrast two analytical techniques: the typical single model approach which uses single estimates for all parameters and a Bayesian approach which incorporates the uncertainties of model choice and parameter estimation. We also consider two classification schemes. The current IUCN scheme, which classifies populations/species according to an X% chance of extinction in Y years (like the Endangered criterion: a 20% chance of extinction in 20 years). Including model and parameter uncertainty results in a probability distribution for each time. For example, a model assuming density dependent dispersal might have a very different probability of extinction in 100 years than a model assuming constant dispersal. Similarly, models using rates of decline of 2% or 4%/year, which may both be plausible given the data, will have different probabilities of extinction for a set period of time. Thus, more appropriate classification criteria would be phrased in terms of assuming a high probability of correctly classifying an endangered population. For example, the endangered criterion might be a 90% chance of correctly classifying a population which was truly endangered (e.g., had a 20% chance of extinction, or greater, in 20 years). We contrast the performance of these criteria and PVA techniques using our simulations where the actual level of risk is known.

Population Viability Assessment Using Monte-Carlo Life-Table Simulations. M. K. TAYLOR*, Fisheries and Wildlife, RWED/GNWT, Iqaluit, NT, Canada, M. KUC, ESSA Technologies, Richmond Hill, Ontario, M. A. OBBARD, Wildlife and Natural Heritage, Ontario MNR, Peterborough, Ontario, B. POND, Wildlife and Natural heritage, Ontario MNR, Peterborough, Ontario, and D. CLUFF, Regional Biologist, RWED/GNWT, Yellowknife, NT, Canada.

Population viability is essentially a question of demographic performance. One approach is demographic simulations based on life-history models of population dynamics. Estimates of population number, sex and age distribution, survival, recruitment, and harvest (if any) may be used in age-structured birth-pulse models to estimate population trend or status, number at some future time, and to explore the demographic consequences of some particular set of circumstances. Models may allow unlimited exponential growth or contain density effect feedback mechanisms. Harvest may be incorporated in a variety of ways, ranging from detailed simulation of actual sex and age selectivity/vulnerability, to simple apportionment of the kill according to the relative abundance of the population sex and age categories. A Windows compatible program named "RISKMAN" (RISK MANagement) was developed for the full range of options described above. RISKMAN provides a stochastic option that is based on the variance of input parameters, and the structure identified by the simulation options that are selected. Monte Carlo techniques are used to generate a distribution of results, and that distribution is used to estimate the variance of summary parameters such as number at some future time or population growth at some future time. RISKMAN utilizes the correct distributions of the population and rate variance estimates to provide accurate estimates of the uncertainty of summary parameters. Parameters may co-vary or be independent, and RISKMAN allows the user to set the correlation to one or zero to bound the co-variance possibilities.

Effects of Inbreeding on Population Viability, Modeling, and Management. DAVID W. TONKYN, Dept. of Biol. Sciences, Clemson Univ., Clemson, SC.

Inbreeding is inevitable in small, isolated populations, and is associated with reduced juvenile survival in many captive populations of endangered vertebrates. However, earlier studies likely underestimated the full effects of inbreeding, both to individuals, in reduced survival and fecundity, and to populations, in increased risk of extinction from low viability, disease, or environmental change. Among other concerns, these studies have not used the full information on living animals, often the most abundant and inbred in the data sets, and have ignored possible additional mortality among adults. Using more powerful statistical models, we have shown that inbreeding reduces survival in endangered Asian leopards and other species, and not only among the young, as was known, but continuously and therefore cumulatively throughout life. As a result, both individual survival to reproductive age and population growth rates decline faster with inbreeding than previously thought, increasing the risk of extinction. In this talk, I shall review these results, discuss specific implications for conservation of the endangered Amur leopard, and then conclude with general remarks on the management of both captive and wild endangered species. If many species exhibit a similar interaction between demography and genetics, this will change our basic understanding of inbreeding, require further integration of demographic and genetic models for accurate PVAs, and perhaps even influence views on breeding strategies for captive populations, and on what constitutes a minimum viable population in nature.

Viability of Feral Horse Populations on Atlantic Coastal Barrier Islands: Implications for Management. H. BRIAN UNDERWOOD, USGS Patuxent Wildlife Research Center, College of Environmental Science & Forestry, Syracuse, NY.

Feral and free-ranging horses have a long and illustrious legacy common to many Atlantic coastal barrier islands. I analyzed the population dynamics of the horses of Assateague Island National Seashore. Using population reconstruction methods and computer simulation of feral horse life-history, I developed a demographic PVA for this population for the expressed purpose of assessing the impacts of fertility control, disease epidemics, and catastrophic storm events on population persistence. Partial pedigrees were constructed to examine lineage-specific effects on population demography and heterozygosity. The reconstructed population varied from a low of 35 horses in 1975 to a high of 171 in 1988. Two-hundred forty-seven observations of survival histories (58 uncensored) of horses demonstrated that reconstructed March population size added significant prognostic value to the prediction of hazards for AINS horses. There were no statistically detectable differences in survival between the sexes and no dramatic differences in the age at first foaling over time. Analysis of the inter-birth interval revealed a strong population size influence, however. There was a pronounced change in the early-age hazards as population size increased, while adult survival changed very little and remained quite high over a wide range in population sizes. The levels of adult mortality due to storms, diseases and accidents, which are impossible to predict, had the most profound effect on population viability. The simulation also suggested that management directed at specific matrilines exerts a disproportionate effect on rate of increase, population viability and heterozygosity.

Mathematical Models as a Basis for Managing Goose Populations in Scotland. MICHAEL B. USHER*, Scottish Natural Heritage, Edinburgh, and ANDREW DOUSE, Scottish Natural Heritage, Edinburgh.

Population viability analyses (PVAs) are being performed on five goose populations in Scotland - the Icelandic population of the greylag goose (Anser anser), the pink-footed goose (Anser brachyrhyncus), the Greenland white-fronted goose (Anser albifrons flavirostris), and the barnacle goose (Branta leucopsis, both Greenland and Svalbard populations). During the last 50 years