Population Viability Analysis

by Steven R. Beissinger (Editor), Dale R. McCullough (Editor)

Overview

As human populations and the resources required to support societies continue to grow, an increasing number of plant and animal species around the world are facing extinction. Given limited time, space, and money, how do we decide which management actions will be most effective to avert extinctions?

In this book, many of the world's leading conservation and population biologists evaluate what has become a key tool in estimating extinction risk and evaluating potential recovery strategies—population viability analysis, or PVA. PVA integrates data on the life history, demography, and genetics of a species with information on environmental variability, using computer models ranging from simple measures of population growth rate to complex spatial simulations, to predict whether a given population will remain viable (i.e., not go extinct) under various management options. A synthetic and objective overview of the latest theoretical and methodological advances, Population Viability Analysis will be crucial reading for conservationists, land managers, and policy makers.

Product Details

ISBN-13:	9780226041780
Publisher:	University of Chicago Press
Publication date:	05/04/2002
Edition description:	1
Pages:	593
Product dimensions:	6.00(w) x 9.00(h) x 1.30(d)

About the Author

Steven R. Beissinger is a professor of conservation biology and chair of the Department of Environmental Science, Policy, and Management at the University of California, Berkeley.

Dale R. McCullough holds the A. Starker Leopold Chair in Wildlife Biology and is professor of wildlife biology in the Department of Environmental Science, Policy, and Management at the University of California, Berkeley.

Read an Excerpt

POPULATION VIABILITY ANALYSIS

---

THE UNIVERSITY OF CHICAGO PRESS
Copyright © 2002 The University of Chicago
All right reserved.
ISBN: 978-0-226-04178-0

---

Chapter One
Population Viability Analysis: Past, Present, Future Steven R. Beissinger

Population viability analysis (PVA) has become a cornerstone of conservation science. It is both a process and a tool that has been used to create theory, to analyze data and project population trends, and to make policy decisions. Viability is a pervasive concept that has contributed almost as much to theory as it has to application. PVA has provided a framework for understanding how stochastic events and processes affect the chances of extinction (Lande, chap. 2 in this volume). The role of stochasticity in extinction has been called the small-population paradigm (Caughley 1994). Few scientists challenge the notion that genetic, demographic, and environmental stochasticity and catastrophes can result in extinction, although the relative role of each process is often debated (Lande 1988, 1993; Allendorf and Ryman, chap. 4 in this volume). Yet it is the way in which we apply viability concepts to evaluating risks and recovery strategies for populations, species, and ecosystems in trouble that has sometimes engendered controversy over both scientific theories and public policy (Mann and Plummer 1999; Shaffer et al., chap. 7 in this volume).

Perhaps these disagreements stem partly from differences in the way that viability processes interact with the processes of recovering endangered species, which Graeme Caughley (1994) so elegantly defined as the declining-population paradigm. The declining-population paradigm consists of determining if there is a population trend, identifying what element of the organism's demography is depressed, determining the environmental or intrinsic factors that cause the depressed rate, and then examining what can be done to correct these limiting factors. In this paradigm, stochasticity is rarely identified as a specific cause of poor reproduction or survival, perhaps with the exception of some genetic factors like inbreeding. The declining-population paradigm and the small-population paradigm offer complementary ways of analyzing the same problem, and they may lead to similar or contradictory conclusions (Boyce, chap. 3 in this volume).

This chapter reviews the past, present, and future of PVA from the perspective of a conservation biologist who works at the interface of academia and management and advocates good science rather than a particular conservation philosophy. In doing so, I try to bridge the worlds of model builders and legacy builders. I begin by examining the history and growth of PVA over the past 20 years and review how it has evolved to date. I next consider the criticisms of PVA and how it has been applied in conservation. I conclude by speculating about how PVA might look in the future.

HISTORY AND GROWTH OF PVA

While the history of PVA is short (Soulé 1987a), the nature of PVA has changed greatly over time. At the beginning of the 1960s, prior to the passage of the first U.S. Endangered Species Act in 1966, the most threatened and publicly visible endangered species in the United States were the California condor (Gymnogyps californianus), whooping crane (Grus americana), tule elk (Cervus elaphus nannodes), and black-footed ferret (Mustela nigripes). It is notable that PVA did not contribute to the recovery of any of these species, except after the fact for the condor and ferret, although deterministic demographic models that estimated the rate of annual population growth had been used to make management recommendations for the whooping crane (Miller and Botkin 1974) and for the California condor (Mertz 1971).

PVA was spawned by a controversy in the late 1970s over the Yellowstone National Park population of grizzly bears (Ursus arctos). The controversy centered on the recommendations made by wildlife researchers John and Frank Craighead and their colleagues regarding the consequences of closing the park dumps to the bear population. Craig-head et al. (1973) used a computer model to show that mortality from the park's approach to problem bears was driving the population to extinction, which led to a review by a National Academy of Sciences committee (Cowan et al. 1974). As a member of the committee, Dale McCullough developed an alternative model that led to differing predictions about the grizzly bear's fate and which was subsequently published (McCullough 1978, 1986). Both models were deterministic models, although McCullough (1986) and Avrin (1976) investigated stochastic recruitment. A model for the grizzly bear that included both demographic and environmental stochasticity was later developed by Mark Shaffer in 1978 as part of his Ph.D. dissertation at Duke University, and to him goes the credit for the first true PVA.

Shaffer's approach (1981) was a new direction in the use of models for conservation, because he developed a stochastic population simulation that incorporated chance events-specifically demographic and environmental stochasticity-and produced extinction probabilities. His model also estimated a minimum viable population size (MVP) by varying the initial number of individuals to find the smallest population size with a 95% chance of remaining extant after the simulated 100-year period (Shaffer and Samson 1985).

The MVP concept was seductive and had arrived at the right moment. The National Forest Management Act of 1976 mandated that harvesting policies and management actions of the U.S. Forest Service (USFS) must sustain viable populations of terrestrial vertebrates. Hal Salwasser and Fred Samson of the USFS were looking for ways to estimate viable population size (Samson 1983; Salwasser et al. 1984; Samson et al. 1985), and they encouraged the agency to incorporate PVA approaches in forest planning. At the same time, viability concepts greatly appealed to decision makers in the U.S. Fish and Wildlife Service who had to implement the Endangered Species Act. They needed a method to quantify risk of extinction to evaluate if a species should be listed as threatened or endangered, to indicate if a federal project adversely affected a species during section 7 consultations, and to determine when a species had recovered and could be proposed for delisting.

Shaffer's MVP paper was followed by the publication of Frankel and Soulé's Conservation and Evolution in 1981. Here for the first time genetic approaches to evaluate the viability of populations were detailed in one place, building on the work of Denniston (1978). The short-term and long-term effects of inbreeding and genetic drift were placed in a framework that clarified their erosive impacts on genetic diversity. Here is where the 50/500 rule for effective population sizes, which had recently been proposed by Franklin (1980) and Soulé (1980), was stated in a manner that became etched into the conservation consciousness, for better or worse, as one of the few rules of thumb that conservation biologists have produced to judge viability.

Soon papers that estimated MVP size began to appear (Schonewald-Cox et al. 1983; Lacava and Hughes 1984; Lehmkuhl 1984; Shaffer and Samson 1985; Menges 1986; Reed et al. 1988). Most of these works employed genetic approaches to determine MVPs by adapting and parameterizing demographic equations for estimating effective population size that had been part of classical population genetics theory. I suspect that demographic applications of PVA lagged behind genetic applications because of the difficulty and time involved in developing Monte Carlo computer applications that were needed to produce MVPs based on stochastic population simulations. Let's not forget that in the mid-1980s the personal computer was just making its appearance on the commercial market, and programming in FORTRAN for mainframe computers using decks of cards could hardly be called user-friendly.

In 1986 the work of Michael Gilpin and Michael Soulé put PVA on firmer footing. Gilpin and Soulé (1986) broadened the definition of PVA when they used it as a heuristic concept to examine the many forces that interact in vortices that can affect the viability of a population, including genetic factors. Although they designated the term "Population Vulnerability Analysis" for this approach, the term never became widely used. However, the idea of PVA as a process of risk analysis emerged, where hazards are identified, risks are considered, and a model is developed in the final step.

The Gilpin and Soulé paper made its debut in 1985 at the Ann Arbor, Michigan, meeting where the Society for Conservation Biology was established. Prior to that meeting, there was a workshop that gathered a few scientists together to work on the MVP concept. The result was a series of papers published in Viable Populations for Conservation (edited by Soulé [1987b]) that established PVA as both a process and a technique. Concurrently, Gilpin and Soulé were working on the first metapopulation PVA model for a threatened water snake (Soulé 1989).

PVA applications grew rapidly in the late 1980s and early 1990s, largely due to two factors. First, zoological parks quickly employed the power of genetic viability concepts to help them manage small populations in captivity to minimize the loss of genetic diversity. Robert Lacy, Jon Ballou, Kathy Ralls, Ulysses Seal, and others developed workshops and computer software programs to assist zoos in managing captive populations. The application of the small-population paradigm has been extremely successful in this regard. Second, demographic PVAs proliferated rapidly with the development of a variety of computer software packages, such as VORTEX, the RAMAS series, and ALEX, that made Monte Carlo population simulations relatively easy to perform (Lindenmayer et al. 1995). These software packages put PVA on the desktop computer of every interested conservation biologist and empowered users who were computer literate but who were not computer programmers. The advent of these computer programs, as much as any other single factor, promoted the rapid proliferation of PVA as a tool for evaluating conservation problems.

As the power of personal computers has grown over the past decade, approaches for modeling populations have changed. Ease of computation led to exploring a greater number of scenarios in simulation models and to more sophisticated sensitivity analyses (Mills and Lindberg, chap. 16 in this volume). Also, in the past 20 years PVA models have become more complex as we began to incorporate spatial processes. Models grew from single populations to metapopulation models that treated demography at the level of a homogeneous patch, then matured to grid models that created spatially explicit addresses for groups of individuals that allowed them to take on demographic characteristics based on the surrounding grid squares, and finally ripened to the current trend of individual-based models that track each individual as it lives, dies, and moves among landscape elements (Beissinger and Westphal 1998).

The increase in model complexity was driven by the need to consider a variety of processes and scales that affect both organisms and management decisions, and by the growth of GIS and remote sensing technologies. The "rediscovery" of Levins's metapopulation concept (1969, 1970) and its application to conservation biology spawned a new spate of theory, more field studies, and sophisticated modeling efforts. Furthermore, Pulliam's work (1988) on sources and sinks reminded us that demography could vary across a landscape. Hanski (chap. 5 in this volume) and Harrison and Ray (chap. 6 in this volume) summarize advances in this field and the applicability of metapopulation models to animal and plant populations, respectively.

With new technologies, conservation biologists now had the tools to make relatively detailed habitat maps. All that was required to build models that were spatially explicit was the ability to construct complex computer programs. The increased complexity of spatially explicit and individual-based models was seductive because they could be applied to specific conservation situations, such as the loss of particular patches to habitat destruction or the role of dispersal corridors (Dunning et al. 1995). However, increasing the complexity of models also greatly inflated the number of model parameters that had to be estimated (Beissinger and Westphal 1998; Ralls et al., chap. 25 in this volume). It is an unusual endangered species for which we have enough data to estimate most parameters in these models. Such species are either inherently easy to study or their welfare is threatened by the extraction of very valuable resources, which results in large financial investments in field research. Even in these cases, data that are critical to model outcomes-such as mortality rates in relation to movement-may not exist.

CURRENT ISSUES IN THE USE OF PVA

Despite the prevalent position of PVA in conservation science, there has been growing concern over how PVA models are used for making conservation decisions (Boyce 1992; Ralls and Taylor 1997; Beissinger and Westphal 1998; Groom and Pascual 1998; Reed et al. 1998). Criticisms have been raised over the manner that inputs, assumptions, and structure of demographic PVA models affect their ability to predict the fate of populations with accuracy and precision. Four dominant causes of errors result in uncertainty in the outcomes predicted by PVA models: (1) poor data, (2) difficulties in parameter estimation, (3) weak ability to validate models, and (4) effects of alternative model structures.

Rarely have detailed field studies with adequate sample sizes been used for developing mean estimates of vital rates, and causes and timing of mortality are seldom known. In animals, survival is often the most difficult vital rate to measure precisely because it must be distinguished from the probability of resighting (Nichols 1992). Although sophisticated mark-recapture statistical techniques have been developed to yield accurate estimators of survival and test for differences among individuals of different ages or stages (Lebreton et al. 1992; White et al., chap. 9 in this volume), they require a minimum of three years of study to estimate probability of resighting and survivorship for a single year, and more years to develop precise estimators when the probability of resighting is low. Unfortunately, lack of good survival data can complicate estimation of population change and extinction rates, because elasticity and sensitivity analyses suggest that population change in long-lived organisms is usually most affected by changes in adult survivorship (Boyce 1992; Silvertown et al. 1993; Pfister 1998; Sæther and Bakke 2000). Estimating survival of adults is much less of a problem in plant studies, since they are nonvagile, but survival of seeds and mating success are correspondingly difficult to measure (Doak et al., chap. 15 in this volume). In addition, survival is difficult to estimate for perennials that experience dormancy. As a way to circumvent this problem, mark-recapture models were recently adapted to estimate dormancy and survival rates for rare perennials (Shefferson et al. 2001).

It is equally challenging to develop robust variance estimators for stochastic models (White et al., chap. 9 in this volume). We still require formal definitions of demographic and environmental stochasticity for age- and stage-structured models, which are well tested and confirmed, to incorporate into PVA software (Sæther and Engen, chap. 10 in this volume). Furthermore, accurate estimates for variance of vital rates may require demographic measurements made over many years to sample the range of environmental variation. Rare events, such as 1-in-50- or 100-year droughts, floods, fires, or storms, likely have large effects on variance estimates and population viability (Ludwig 1996, 1999). Long-term studies have shown that estimated variance in population size does not begin to asymptote until after 8 to 20 years, if at all (Pimm 1991). Use of data from short-term studies will usually underestimate the variance in vital rates. However, the effect of short study periods may be offset because estimates of variance derived from field studies include sampling error, which results in an overestimate of variance (Beissinger and Westphal 1998; White et al., chap. 9 in this volume).

(Continues...)

---

Excerpted from POPULATION VIABILITY ANALYSIS Copyright © 2002 by The University of Chicago. Excerpted by permission.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.

---

Table of Contents

Foreword: Raising the Bar
Michael E. Soulé
Editors' Preface
Part I. Overview of Population Viability Analysis
1. Population Viability Analysis: Past, Present, Future
Steven R. Beissinger
2. Incorporating Stochasticity in Population Viability Analysis
Russell Lande
3. Reconciling the Small-Population and Declining-Population Paradigms
Mark S. Boyce
4. The Role of Genetics in Population Viability Analysis
Fred W. Allendorf and Nils Ryman
5. Metapopulations of Animals in Highly Fragmented Landscapes and Population Viability Analysis
Ilkka Hanski
6. Plant Population Viability and Metapopulation-Level Processes
Susan Harrison and Chris Ray
7. Population Viability Analysis and Conservation Policy
Mark Shaffer, Laura Hood Watchman, William J. Snape III, and Ingrid K. Latchis
Part II. Issues in the Parameterization and Construction of PVA Models
8. Definition and Estimation of Effective Population Size in the Conservation of Endangered Species
Robin S. Waples
9. Estimating Parameters of PVA Models from Data on Marked Animals
Gary C. White, Alan B. Franklin, and Tanya M. Shenk
10. Including Uncertainties in Population Viability Analysis Using Population Prediction Intervals
Bernt-Erik Sæther and Steinar Engen
11. Bayesian Population Viability Analysis
Paul R. Wade
12. Incorporating Uncertainty in Population Viability Analyses for the Purpose of Classifying Species by Risk
Barbara L. Taylor, Paul R. Wade, Uma Ramakrishnan, Michael Gilpin, and H. Resit Akçakaya
Part III. Integrating Theory and Practice in the Use of Population Viability Analysis
13. How Good Are PVA Models? Testing Their Predictions with Experimental Data on the Brine Shrimp
Gary E. Belovsky, Chad Mellison, Chad Larson, and Peter A. Van Zandt
14. Evolution of Population Viability Assessments for the Florida Panther: A Multiperspective Approach
David S. Maehr, Robert C. Lacy, E. Darrell Land, Oron L. Bass Jr., and Thomas S. Hoctor
15. Population Viability Analysis for Plants: Understanding the Demographic Consequences of Seed Banks for Population Health
Daniel F. Doak, Diane Thomson, and Erik S. Jules
16. Sensitivity Analysis to Evaluate the Consequences of Conservation Actions
L. Scott Mills and Mark S. Lindberg
17. Application of Molecular Genetics to Conservation: New Issues and Examples
Philip W. Hedrick
18. Pedigree Analyses in Wild Populations
Susan M. Haig and Jonathan D. Ballou
19. Rangewide Risks to Large Populations: The Cape Sable Sparrow as a Case History
Stuart L. Pimm and Oron L. Bass Jr.
20. Population Viability Analysis, Management, and Conservation Planning at Large Scales
Fred B. Samson
Part IV. The Future of Population Viability Analysis
21. Predictive Bayesian Population Viability Analysis: A Logic for Listing Criteria, Delisting Criteria, and Recovery Plans
Daniel Goodman
22. Decision Theory for Population Viability Analysis
Hugh P. Possingham, David B. Lindenmayer, and Geoffrey N. Tuck
23. Incorporating Human Populations and Activities into Population Viability Analysis
Robert C. Lacy and Philip S. Miller
24. Fitting Population Viability Analysis into Adaptive Management
Donald Ludwig and Carl J. Walters
25. Guidelines for Using Population Viability Analysis in Endangered-Species Management
Katherine Ralls, Steven R. Beissinger, and Jean Fitts Cochrane
About the Contributors
Index