Back to Microsoft Word 11.0.6568;SOCIETY FOR CONSERVATION BIOLOGY NEWSLETTER - Volume 13, Issue 3 - August 2006

Up to Table of Contents

Ahead to SCB DECIDES TO OFFSET ITS CARBON IMPACT

THE LOGIC OF GOOD DECISIONS: LEARNING FROM POPULATION VIABILITY ANALYSIS

BY MARK BURGMAN

• Learning from PVA
• Conclusions
• Acknowledgements
• References

Accurate predictions improve decisions. However, it may be just as important that predictions are comprehensible, persuasive, feasible, coherent, or authoritative. Effective social outcomes involve group decisions and usually depend on participatory decision making processes (O'Brien 2000, Pielke 2003). Models are important for conservation decision-making because they combine our current understanding of ecological systems with objective functions that measure the value of management alternatives. Stochastic models of the dynamics and persistence of species (population viability analyses, Reed et al. 2002) are important because of the pervasiveness of natural variation and lack of ecological knowledge. The new challenge in conservation biology is to develop frameworks that harness tools such as PVA to improve decision-making (Reed et al. 2002, Beissinger et al. 2006).

Steele et al. (in press) outlined a number of substantial issues in decision-making relevant to conservation biology. The objectively "correct" decision might not please any members of a group. Conversely, a popular decision may fail to be the correct decision. The conditions under which happy decision makers and correct decisions coincide may be enhanced by formal consensus methods. However, Steele et al. emphasised that consensus is futile when no fact of the matter exists.

In most practical circumstances, sufficient data are unavailable for accurate predictions. Typically in conservation biology, we lack understanding of population dynamics, parameters, spatial processes, distributional shapes, and dependencies. One attribute of good decisions is that they are robust to uncertainty. They provide for the possibility that models and data are wrong. For example, Regan et al. (2005) concluded that the most robust management option provided at least some benefits for a species, irrespective of the (uncertain) causes of decline. Explicit, coherent information about uncertainty gives decision-makers the opportunity to evaluate their attitudes to risk. For instance, Halpern et al. (2006) concluded that explicit modeling of uncertainty in marine planning allowed managers to quantify costs and benefits, and to communicate to management challenges to stakeholders.

Population viability analyses have been criticized because confidence intervals for parameters and risk metrics are large. Models may be unnecessary, unhelpful, wasteful, or misleading because predictions are difficult (for discussion, see Reed et al. 2002, Brook et al. 2002, Beissinger et al. 2005 and references therein). Many criticisms of PVA are valid. Often, there are too few data to answer a question. Functions and parameters may over or underestimate future population sizes. Bounds on most predictions for single species are disturbingly large. The variability represented in most PVAs underestimates the full extent of natural variation and lack of knowledge.

Critiques of PVA have stimulated tests of its strengths and weaknesses, employing theoretical simulations, retrospective and cross-validation studies, and empirical tests of predictive accuracy (Brook et al. 2000, Ellner et al. 2002, Lindenmayer et al 2003). They have explored biases and missing data (Holmes and Fagan 2003, McCarthy et al. 2004), the relative accuracy of predicted ranks of management options (Ellner and Fieberg 2003, McCarthy et al. 2003, Lotts et al. 2004, Holmes et al. 2005), and have decomposed predictive uncertainty into its sources (Ellner and Fieberg 2003, Wiegand et al. 2004).

New analytical techniques are evolving that take a broad range of uncertainties, propagate them through chains of calculations, and provide platforms for formal decision analysis (e.g., Ben-Haim 2001, Ferson 2002, Possingham et al. 2002, Dorazio and Johnson 2003). They provide decision-makers with a more transparent appreciation of what is possible to predict.

Learning from PVA

Prediction is a part of the process of quantitative modeling. The heuristic benefits (Brook et al. 2002) accrue from the full process of building, fitting, predicting, testing, and revising the models. Of these, arguably, model building is the most valuable because it contributes the most to learning (Starfield 1997).

The value of any of the steps in Figure 1 depends on context. Discrepancies, logical flaws, and mismatched units in conceptual models are often illuminated when they are translated into equations. Model building is the simplest way of finding and reconciling the linguistic ambiguities and vague concepts that plague informal risk assessments (Regan et al. 2002). It may be enough to establish data collection and research priorities from a sensitivity analysis. The ranks of management options may be sufficient basis for managing a species.

Even when uncertainties are large, PVAs can guide species management, set research priorities, establish recovery criteria, support monitoring programs, and characterize the risks faced by populations (see Staples et al. 2005, Beissinger et al. 2006). For example, Yamada et al. (2004) built a model for Sindh Ibex in Khirthar National Park, Pakistan, based largely on collateral data and allometric relationships, together with historical records and expert judgement. Information on dependencies between vital rates (within populations and in space) could only be inferred from rainfall data. Dependencies can strongly affect extinction probabilities (Burgman 2005). However, a sensitivity analysis indicated that dependencies between vital rates and in space, for this species, contributed a relatively modest amount to uncertainty. Further, a decision about whether it was safe to harvest 25 adult males from an adjacent game reserve would not change, irrespective of knowledge about these parameters.

Work over the last 20 years has led to a realization that much uncertainty is irreducible (inherent, natural variation) and that some theoretically reducible uncertainty is not within our grasp (Pielke 2003). More knowledge may increase uncertainty because better models enhance appreciation of its magnitude. Without such insight, decisions often are unrealistically optimistic (Burgman 2005).

Conclusions

The easy solution for conservation decision-making under uncertainty is to turn to expert judgment or to a single model-based prediction, thereby assuming a cloak of unassailable and unverifiable scientific authority to produce persuasive rationalizations. Unfortunately, many alternative scenarios may unfold and both experts and models are characteristically optimistic and narrow in their appreciation of uncertainty (Burgman 2005). Alternative views of a system may be coherent and consistent with data and theory. Conservation needs decisions that provide satisfactory utilities for all stakeholders, irrespective of which model for the future is true.

Beissinger et al. (2006) concluded that the best conservation decisions will occur where field biologists, modelers, statisticians, and managers cooperate. Good decision-making makes use of the full power of explicit modeling to explore data and the status of knowledge relevant to a decision, and then gives uncertainty full credibility by finding a solution that is maximally robust to uncertainty. The task remains in conservation biology to embrace tools for uncertainty analysis in ways such that the models are subordinate to the quality of the data, the people who carry the burden of the risks have a place and a voice, and field biology, modeling, and management planning contribute to decisions.

Acknowledgements

I am grateful to Steve Beissinger, Mick McCarthy, and Jane Elith for their comments.

Mark Burgman received a 2006 Distinguished Service Award from SCB for his contributions to the analysis of problems and practical management in conservation biology through his unique perspective of incorporating risk and uncertainty in the development of management options for conservation.

References

Beissinger, S.R., J.R. Walters, D.G. Catanzaro, K.G. Smith, J.B. Dunning, S.M. Haig, B.R. Noon, and B.M. Stith. 2006. Ornithological Monographs 59:1-56.

Ben-Haim, Y. 2001. Information-gap decision theory: decisions under severe uncertainty. Academic Press, San Diego.

Brook, B.W., J.J. O'Grady, A.P. Chapman, M.A. Burgman, H.R. Akçakaya, and R. Frankham. 2000. Nature 404:385-387.

Brook, B.W., M.A. Burgman, H.R. Akçakaya, J.J. O'Grady, and R. Frankham. 2002. Conservation Biology 16:262-263.

Burgman, M.A. 2005. Risks and decisions for conservation and environmental management. Cambridge University Press.

Dorazio, R.M. and F.A. Johnson. 2003. Ecological Applications 13:556-563.

Ellner, S., J. Fieberg, D. Ludwig, and C. Wilcox. 2002. Conservation Biology 16:258-261.

Ellner, S.P. and J. Fieberg. 2003. Ecology 84:1359-1369.

Ferson, S. 2002. RAMAS RiskCalc Version 4.0. Software: risk assessment with uncertain numbers. Lewis, Boca Raton, Florida.

Halpern, B.S., H.M. Regan, H.P. Possingham, and M.A. McCarthy. 2006. Ecology Letters 9:2-11.

Holmes, E.E. and J.J. Fagan. 2003. Ecology 83:2379-2386.

Holmes, E.E., W.F. Fagan, J.J. Rango, A. Folarin, J.A. Sorensen, J.E. Lippe, and N.E. McIntyre. 2005. Cross validation of quasi-extinction risks from real time series: an examination of diffusion approximation methods. U.S. Department of Commerce, NOAA Technical Memo. NMFS-NWFSC-67.

Lindenmayer, D.B., H.P. Possingham, R.C. Lacy, M.A. McCarthy, and M.L. Pope. 2003. Ecology Letters 6:41-47.

Lotts, K.C., T.A. Waite, and J.A. Vucetich. 2004. Conservation Biology18:1224-1232.
McCarthy, M.A., S.J. Andelman, and H.P. Possingham. 2003. Conservation Biology 17:982-989.

McCarthy, M.A., D. Keith, J. Tietjen, M.A. Burgman, M. Maunder, L. Master, B.W. Brook, G. Mace, H.P. Possingham, R. Medellin, S. Andelman, H.M. Regan, T. Regan, and M. Ruckelshaus. 2004. Acta Oecologica 26:67-74.

O'Brien, M. 2000. Making better environmental decisions: an alternative to risk assessment. MIT Press, Cambridge, Massachusetts.

Pielke, R.A. 2003. Pages 111-135 in C. Canham, J. Cole, and W.K. Lauenroth, editors. Models in ecosystem science. Princeton University Press, New Jersey.

Possingham, H.P., D.B. Lindenmayer, and G.N. Tuck. 2002. Pages 470-489 in S.R. Beissinger and D.R. McCullough, editors. Population viability analysis. University of Chicago Press, Chicago, Illinois.

Reed, J.M., L.S. Mills, J.B. Dunning, E.S. Menges, K.S. McKelvey, R. Frye, S.R. Beissinger, M.-C. Anstett, and P. Miller. 2002. Conservation Biology 16:7-19.

Regan, H.M., M. Colyvan, and M.A. Burgman. 2002. Ecological Applications 12:618-628.

Regan, H.M., Y. Ben-Haim, B. Langford, W.G. Wilson, P. Lundberg, S.J. Andelman, and M.A. Burgman. 2005. Ecological Applications 15:1471-1477.

Staples, D.F., M.L. Tapr, and B.B. Shepard. 2005. Conservation Biology 19:1908-1916.

Starfield, A.M. 1997. Journal of Wildlife Management 61:261-270.

Steele, K., H.M. Regan, M. Colyvan, and M.A. Burgman. In press. Right decisions or happy decision-makers? Social Epistemology.

Wiegand, T., E. Revilla, and F. Knauer. 2004. Biodiversity and Conservation 13:53-78.

Yamada, K., M. Ansari, R. Harrington, D. Morgan, and M.A. Burgman. 2004. Pages 469-481 in H.R. Açkakaya, M.A. Burgman, O. Kindvall, P. Sjogren-Gulve, J. Hatfield, and M. McCarthy, editors. Species conservation and management. Oxford University Press, New York.

Back to Microsoft Word 11.0.6568;SOCIETY FOR CONSERVATION BIOLOGY NEWSLETTER - Volume 13, Issue 3 - August 2006

Up to Table of Contents

Ahead to SCB DECIDES TO OFFSET ITS CARBON IMPACT