Applications and Challenges for Adaptive Management

University of Nebraska - Lincoln DigitalCommons@University of Nebraska - Lincoln

Dissertations & Theses in Natural Resources Natural Resources, School of

4-2011

Follow this and additional works at: https://digitalcommons.unl.edu/natresdiss

Part of the Natural Resources and Conservation Commons

McFadden, Jamie E., "Applications and Challenges for Adaptive Management" (2011). Dissertations & Theses in Natural Resources. 22. https://digitalcommons.unl.edu/natresdiss/22

This Article is brought to you for free and open access by the Natural Resources, School of at DigitalCommons@University of Nebraska - Lincoln. It has been accepted for inclusion in Dissertations & Theses in Natural Resources by an authorized administrator of DigitalCommons@University of Nebraska - Lincoln.

APPLICATIONS AND CHALLENGES FOR ADAPTIVE MANAGEMENT

Jamie E. McFadden

A THESIS

Presented to the Faculty of

The Graduate College at the University of Nebraska

In Partial Fulfillment of Requirements

For the Degree of Master of Science

Major: Natural Resource Sciences

Under the Supervision of Professor Andrew J. Tyre

Lincoln, Nebraska

April 2011

APPLICATIONS AND CHALLENGES FOR ADAPTIVE MANAGEMENT

Jamie E. McFadden, M.S.

University of Nebraska, 2011

Advisor: Andrew J. Tyre

Adaptive management is becoming an increasingly popular management-decision tool within the scientific community. The application of adaptive management is appropriate for complicated natural-resources management problems high in uncertainty.

Two primary schools of thought have developed that may yield varying levels of success as they primarily differ in stakeholder involvement and model complexity. I evaluated peer-reviewed literature that incorporated adaptive management to identify components of successful adaptive management plans and to make comparisons between the two schools of thought. Identifying the elements of successful adaptive management is advantageous to natural-resources managers, such as those managing the Platte River,

Nebraska. The Platte River is a complicated ecosystem where management decisions affect endangered and threatened species such as the Interior Least Tern (Sternula antillarum athalassos) and Piping Plover (Charadius melodus). Because high uncertainty is associated with these species’ responses to habitat restoration and other resource uses and management efforts differ between the lower Platte River (LPR) and

the central Platte River (CPR), I developed quantitative applications for each section. For terns and plovers on the LPR, I developed a population model that estimates population characteristics for on-channel and off-channel habitat. Model results suggest that population sizes respond similarly for short-term simulations, but differ for long-term scenarios. The ability of this quantitative model to adapt to new information makes it

ideal for projecting management implications within an adaptive management context.

As the CPR is further along in the adaptive management process, I developed a multi- model analysis of 10 models based on simulated data to simplify hypotheses and prioritize management needs. Model results suggest that in not accounting for overdispersion in the data leads to a greater probability of concluding a false relationship when the parameter effect sizes are close to 0. Utilizing statistical models to evaluate management consequences through an iterative decision-making process allows for continuous model improvements based directly on monitoring data. The process of evaluating effects of ecological factors is helpful in setting and prioritizing objectives and implementing actions for adaptively managing complicated ecosystems.

DEDICATION

To my mother and brother: for your endless love and support in all my adventures and for sharing in my awe of God’s creation.

ACKNOWLEDGEMENTS

Immeasurable thanks to my advisor, Dr. Andrew J. Tyre, for giving me the opportunity to pursue a master’s degree and for his unending guidance, support, and patience. I am forever grateful for the countless leadership experiences that Dr. Tyre provided for me. I could not ask for a better advisor! I thank Dr. Craig Allen and Dr.

Tala Awada for their time and encouragement as members of my graduate committee.

For sharing their years of experience in structured decision-making strategies, adaptive management, and ecology with me, I thank Drs. Tyre, Wedin, and Allen, as well as the folks from the National Conservation Training Center.

This thesis is a result of collaboration of many individuals. I am thankful for the support of a Center for Great Plains Studies Grant and funding by the Tern and Plover

Conservation Partnership, the Nebraska Game and Parks Commission, Headwaters

Corporation, and the School of Natural Resources at the University of Nebraska-Lincoln.

Specifically, I thank Mary Bomberger Brown of the Tern and Plover Conservation

Partnership; Joel Jorgensen of the Nebraska Game and Parks Commission; and Chad

Smith of Headwater’s Corporation for their support and collaboration. They aided in my understanding of various data gaps and management needs that were essential to guiding

this research. I greatly appreciate their expertise and patience.

I would like to thank my lab mates who provided endless insight and for conversations that were helpful in so many ways. I am grateful for the Tyre-Tenhumberg and Pegg families and Susan Vosler who provided wonderful support and guidance on life’s challenges. Special thanks to all my family, especially my mother, Brenda

McFadden, and my brother, Ken McFadden. I would not be who I am or where I am today without their love and support throughout my life. I am especially thankful for the continuous encouragement from Tim Hiller.

vii

TABLE OF CONTENTS

ABSTRACT ...... ii DEDICATION ...... iv ACKNOWLEDGEMENTS ...... v TABLE OF CONTENTS ...... vii LIST OF TABLES ...... ix LIST OF FIGURES ...... xi Chapter 1: OVERVIEW ...... 1 LITERATURE CITED ...... 9 Chapter 2: EVALUATING THE EFFICACY OF ADAPTIVE MANAGEMENT APPROACHES: IS THERE A FORMULA FOR SUCCESS? ...... 15 Abstract ...... 15 INTRODUCTION ...... 17 METHODS ...... 25 RESULTS ...... 28 DISCUSSION ...... 32 LITERATURE CITED ...... 36 Chapter 3: PROJECTING POPULATION RESPONSES TO MANAGEMENT ON THE LOWER PLATTE RIVER, NEBRASKA, FOR LEAST TERNS AND PIPING PLOVERS WITH A QUANTITATIVE MODEL ...... 44 Abstract ...... 44 INTRODUCTION ...... 46 METHODS ...... 49

vii STUDY AREA ...... 49

NESTING HABITAT TYPES ...... 50 CONCEPTUAL MODEL ...... 50 POPULATION MODEL ...... 52 Population Dynamics ...... 52 Parameter Input ...... 57 HABITAT MODEL ...... 58 SCENARIO DESCRIPTION...... 63

viii

SENSITIVITY ANALYSIS ...... 65 RESULTS ...... 65 SCENARIO RESULTS ...... 65 SENSITIVITY ANALYSIS ...... 73 DISCUSSION ...... 76 MODEL APPLICATIONS ...... 78 NEXT STEPS ...... 79 MODEL CONCLUSIONS ...... 81 LITERATURE CITED ...... 83 Chapter 4: ASSESSING COMPETING RESEARCH HYPOTHESES FOR EXPERIMENTATION THROUGH ADAPTIVE MANAGEMENT ON THE CENTRAL PLATTE RIVER, NEBRASKA ...... 90 Abstract ...... 90 INTRODUCTION ...... 92 METHODS ...... 97 RESULTS ...... 103 AIC MODEL SELECTION ...... 104 MODEL RANKING ...... 107 RECOMMENDED SAMPLE SIZE ...... 108 DISCUSSION ...... 112 LITERATURE CITED ...... 116 Chapter 5: CONCLUSIONS ...... 122 LITERATURE CITED ...... 126

Appendix A: PEER-REVIEWED SCIENTIFIC LITERATURE USED IN THE viii

ANLAYSIS AND DISTRIBUTED BY SCHOOL OF THOUGHT FOLLOWED BY SUCCESS CATEGORY ...... 128 Appendix B: EXTENDED AKAIKE’S INFORMATION CRITERION RESULTS ... 142

LIST OF TABLES

Chapter 2. Evaluating the Efficacy of Adaptive Management Approaches: Is there

a Formula for Success?

Table 1. Comparison of five selected decision-making methods within the adaptive

management literature including Gunderson’s et al. (1995) Adaptive

Environmental Assessment and Management (AEAM), Possingham’s (2000)

Structured Decision Making (SDM), Collaborative Adaptive Management

Network (CAMNet; 2004), Department of Interior (DOI) Adaptive Management

(AM) Protocol from the DOI AM Technical Guide (Williams et al., 2007), and

Foundations of Success (FOS) with the Sustainable Ecosystem’s Institute (SEI,

2007). Comparison criteria include nine adaptive management related variables

found from adaptive management literature along where variables were ordered

(i.e. Order of Variables) according to their sequence within each decision-making

method……………………………………………………………………………22

Chapter 3. Projecting Population Responses to Management on the Lower Platte

River, Nebraska, for Least Terns and Piping Plovers with a Quantitative

Model

Table 1. Input parameters and associated coefficient of variations and standard errors for

tern and plover population model that were obtained from current literature,

research and monitoring estimates, or best estimates from specialists.……..…57

Table 2. Thirteen scenarios developed for model demonstration purposes only that vary

by the loss rate of the specified habitat type area.……………………………..64

Table 3. Sensitivity analysis results for Least Tern data where the fitted linear model

estimates show an overall p-value of <2.2e-16 with r2 = 0.955 and F =

1011……………………………………….……………………………….…..…75

Table 4. Sensitivity analysis for Piping Plover data where the fitted linear model

estimates show an overall p-value of <2.2e-16 with r2 = 0.786 and F = 240.9….75

Chapter 4. Assessing Competing Research Hypotheses for Experimentation

through Adaptive Management on the Central Platte River, Nebraska

Table 1. Models derived from research hypotheses developed by the Platte River

Recovery Implementation Program used in multi-model analysis with Least Tern

adult numbers as the response variable.……………………………...…………..97

Table 2. Parameter effect sizes for each simulation scenario. Assigned values are not

based on real data. Distributions are identical for each parameter (i.e. identical

means and standard deviations)…………………………………………...……..99

Table 3. Results for generalized linear model selection for a single iteration the Strong

Effect scenario computed with a Poisson distribution and Akaike’s Information

Criterion, a Poisson distribution and Quasi-AIC, and a Negative Binomial

distribution and AIC using Program R for model evaluation………..…………105

Appendix B. Extended Akaike’s Information Criterion Results

Table 1. Average Akaike’s Information Criterion (AIC) components, estimated

frequencies of each model ranked as the best AIC model, and mean AIC model

ranking over 100 iterations for each scenario and analysis method……………142

LIST OF FIGURES

Chapter 2. Evaluating the Efficacy of Adaptive Management Approaches: Is there

a Formula for Success?

Figure 1. A comparison of the two dominant adaptive management schools of thought:

the Resilience-Experimentalist Adaptive Management School and the Decision-

Theoretic Adaptive Management School………...………………………………21

Figure 2. Number of scientific articles that reference the term adaptive management (n =

96; not including the Mention articles) from eight scientific journals by year from

2000 to 2009. Linear regression suggested a slope of 0.92 and r2 = 0.56………30

Figure 3. The percentage of scientific articles in the suggest category over time from

2000–2009. Linear regression suggested a slope of 0.0346 and r2 = 0.2442…...30

Figure 4. The percentage of scientific articles by success category for the Resilience-

Experimentalist School of Thought (20% (n = 9) to Theory, 39% (n = 18) to

Suggest, 30% (n = 14) to Framework, and 11% (n = 5) to Implement), the

Decision-Theoretic School of Thought (0% (n = 2) to Theory, 26% (n = 6) to

Suggest, 35% (n = 8) to Framework, and 30% (n = 7) to Implement), and other

(25% (n = 6) to Theory, 67% (n = 16) to Suggest, 0% (n = 1) to Framework, and

0% (n = 1) to Implement)………………………………………..………………31

Figure 5. The percentage of scientific articles with some reference of the term adaptive

management (not including the Mention articles) from eight scientific journals by

year from 2000 to 2009 categorized by two adaptive management schools, the

Resilience-Experimentalist with r2 = 0.00; m = 0.0013; n = 10 (41 total articles)

xii

and the Decision-Theoretic with r2 = 0.67; m = 0.0232; n = 10 (24 total

articles)…………………………………………………………………………...32

Chapter 3. Projecting Population Responses to Management on the Lower Platte

River, Nebraska, for Least Terns and Piping Plovers with a Quantitative Model

Figure 1. Conceptual model depicting the interactions between the habitat model and

Piping Plover population model. Bold square boxes display state variables and

round-edged boxes show processes between state variables indicated through

arrows. The habitat loss rates include changes in erosion and vegetation

processes for sandbar area and include changes in development and mining

operations that affect sandpit area……………………………………………...... 51

Figure 2. Conceptual model depicting the interactions between the habitat model and

Least Tern population model. Bold square boxes display state variables and

round-edged boxes show processes between state variables indicated through

arrows. The habitat loss rates include changes in erosion and vegetation

processes for sandbar area and include changes in development and mining

operations that affect sandpit area………………………………………………..51

xii

Figure 3. The historical mean daily discharge data of the lower Platte River from USGS

Gauge No. 0679600 from 04/01/1949 through 12/31/2009 at North Bend,

Nebraska.……………………………………………………………………...... 60

Figure 4. The available sandbar habitat area on the lower Platte River as a function of

discharge from Parham (2007) applied to the USGS Gauge No. 0679600 daily

xiii

mean discharge data and habitat area data obtained from Nebraska Game and

Parks Commission…………………...... 60

Figure 5. Selected peak mean daily discharge on the lower Platte River during the month

of May from USGS Gauge No. 0679600 from 04/01/1949–12/31/2009 at North

Bend, Nebraska.……………………………………………………………….....61

Figure 6. Results of the long-term model scenarios show Least Tern (long dash line) and

Piping Plover (short dash line) median population responses to differences in

habitat loss rates (solid line) over 92 years. Habitat loss rates for Status Quo

Long-term Scenario are 0.00 for both habitat types (A) and are 0.10 for both

habitat types in the Total Loss Long-term Scenario (B)…..……………………..66

Figure 7. Results for the short-term model scenarios show Least Tern (long dash line)

and Piping Plover (short dash line) median population responses to various

differences in available habitat area (solid line). Habitat loss rates for Sandbar

Gain & Sandpit Loss Scenario are -0.05 for sandbars and 0.05 for sandpits (A),

and are 0.05 for sandbars and -0.05 for sandpits in the Sandbar Loss & Sandpit

Gain Scenario (B)……………………………………………………………..…67

Figure 8. Results for the short-term model scenarios showing Least Tern (long dash line) xiii

and Piping Plover (short dash line) median population responses to differences in

habitat loss rates (solid line). Habitat loss rates for Status Quo Scenario are 0.00

for both habitat types (A), are -0.05 for both habitat types in the Total Gain

Scenario (B), and are 0.10 for both habitat types in the Total Loss Scenario

(C)………………………………………………………………………………68

xiv

Figure 9. Results for the short-term model scenarios show Least Tern (long dash line)

and Piping Plover (short dash line) median population responses to various

decreases in habitat area (solid line). Habitat loss rates for Sandpit Loss Scenario

are 0.00 for sandbars and 0.05 for sandpits (A), are 0.00 for sandbars and 0.10 for

sandpits in the Extreme Sandpit Loss Scenario (B), are 0.05 for sandbars and 0.00

for sandpits in the Sandbar Loss Scenario (C), and are 0.10 for sandbars and 0.05

for sandpits in the Extreme Sandbar Loss Scenario (D)…………………..……..69

Figure 10. Results for the short-term model scenarios show Least Tern (long dash line)

and Piping Plover (short dash line) median population responses to increases in

habitat area (solid line). Habitat loss rates for Sandpit Gain Scenario are 0.00 for

sandbars and -0.05 for sandpits (A) and are -0.05 for sandbars in the Sandbar Gain

Scenario (B)………………………………………………………….………..…71

Figure 11. Model results showing the median (horizontal line in center for each box),

25th and 75th percentiles (ends of boxes), 10th and 90th percentiles (vertical lines,

and outliers (open circles) over all 1,000 runs for each scenario at the conclusion

of the model time frame. Initial starting habitat area varies over all scenarios with

an average of 725±5.49 hectares……………………………………………...….72 xiv

Figure 12. The proportion of runs during the final year of each scenario that show a

decline in habitat area and population size. Shown are short-term scenarios (1-

11), long-term scenarios (12 and 13), habitat area (black bars), plover population

size (dark gray bars), and tern population size (light gray bars)……………...….73

Figure 13. Model sensitivity output for Least Tern population, where much variation

between the population size (i.e., Log Number of Adults) and the survival rates

are attributed to natural variation and that input parameters vary throughout the

time frame of each model run. The population model suggests that population

size increases as survival rates of after second year birds (A), fledglings (B), and

second year birds (C) increase……………………………………………….…..74

Chapter 4. Assessing Competing Research Hypotheses for Experimentation through Adaptive Management on the Central Platte River, Nebraska

Figure 1. Underlying assumption for simulated data scenarios where bare sandpit area

has a negative effect on the number of tern adults nesting on bare river sand.

Shown is the assumption specific to parameter effect sizes in the Strong Effect

data scenario where white indicates an increase in tern population size and red

indicates a decrease in tern population size…………………………………….100

Figure 2. Simulated results of the tern population size in response to bare river sand area

for a single iteration of each simulation scenarios, including the No Effect (A),

Tapering Baseline (B), Similar Effects (C), Strong Effects (D), and Large

Difference (E)………………………………………………………………..…104

Figure 3. Estimated frequencies for each model ranked as the best Akaike’s Information

Criterion (AIC) model out of 100 iterations using Ivlev’s Index for Electivity.

Shown are results for the No Effect (A), Tapering Baseline (B), Similar Effects

(C), Strong Effects (D), and Large Difference (E) data scenarios using the Poisson

distribution with AIC (orange cirlces), Poisson distribution with Quasi-AIC

xvi

(green squares), and the Negative Binomial distribution with AIC (blue triangles)

analysis methods. The model sets were consistent for each scenario and analysis

method……………………………………………………………………….….109

Figure 4. Cumulative probability results for the model weight diversity (Hs) for each

simulated data scenario over 100 iterations using the Poisson distribution and AIC

analysis method as an example. Shown is the Hs for the No Effect Scenario (red

line), Tapering Baseline scenario (light blue line), Similar Effects scenario (green

line), Strong Effects scenario (black line), and Large Difference scenario (light

blue)………………………………………………………...………………..…110

Figure 5. Results for power analysis of the No Effect (A), Tapering Baseline (B), Similar

Effects (C), Strong Effects (D), and Large Difference (E) scenarios showing

model weight diversity (Hs) as a function of sample size (N). Shown is the mean

(horizontal line in center for each box), 25th and 75th percentiles (ends of boxes),

10th and 90th percentiles (vertical lines), and outliers (open circles) over 100

iterations for each sample size test…………….……………………………..…111

xvi

Chapter 1: OVERVIEW

Natural-resources managers are faced with value-laden decisions high in complexity, risk, and uncertainty (Levin, 1999; Gunderson and Holling, 2002; Berkes,

2004). When decisions must be made regardless of the level of knowledge or uncertainty, the application of conventional research methods is often insufficient to support effective decision-making under these circumstances. There is a critical need to improve how research is incorporated into management decisions where uncertainty places limitations on contributions of science (Reynolds et al., 1996; Lee and Bradshaw,

1998; Berg et al., 1999; Robertson and Hull, 2001). In other disciplines, such as business and economics, complex decisions involving risk are often approached using structured decision-making (SDM), described by proponents as "a formalization of common sense for situations too complicated for the informal use of common sense" (Keeney, 1982, p.

806). Although such formal decision-making skills may be underdeveloped by natural- resources managers, the use of SDM is becoming more prevalent within the field of natural resources (Gregory and Keeney, 2002; Conroy et al., 2008; Gregory and Long,

2009). Natural-resources managers can learn through an ongoing process of implementing various management actions, monitoring management outcomes, and

updating ecological models by comparing actual outcomes with expected outcomes by repeating decisions within an SDM approach (Hilborn and Walters, 1981; Walters, 1986;

Williams, 1996; Carpenter, 2002; Johnson et al., 2002). This ongoing (i.e., iterative) learning process, learning by doing, is known as adaptive management.

Throughout the development of natural-resources management, research, and monitoring, natural-resources managers have experienced numerous advancements in monitoring and research methods. However, a need for further improvement and development of the tools for decision making that integrate an active-learning process remains (Walters and Holling, 1990; Walters, 1997). Adaptive management is becoming an increasingly popular concept and has developed within several governmental agencies, resulting in varying definitions of the process. However, there are commonalities amongst the various agencies regarding the adaptive management process, including establishing an iterative process that involves sharing of responsibilities and decision- making among managers, biologists, and stakeholders (Ruitenbeek and Cartier, 2001).

These decision-makers collaborate to develop management plans that allow for analyses of large-scale ecosystem problems through implementing various management actions based on appropriate measurable objectives (Walters, 1997; Holling, 2001; Hughes et al.,

2007). Managers can also apply learning by doing to the particular adaptive management approaches that have been implemented (Johnson, 2006; Runge et. al, 2006; Williams et. al, 2007). In other words, it is necessary to “adaptively manage” the field of adaptive management by testing different decision-making and modeling approaches, monitoring

these management outcomes, and changing our practices to deliver better management outcomes.

While individual approaches to adaptive management are often promoted, there is no comparative overview of different adaptive management approaches (i.e., schools of thought). Scientific literature acknowledges that for the successful application of

3 adaptive management, there must be a cumulative experience of the process through building a thorough understanding of the various elements (Gerber et al., 2007). Overall, with multiple approaches emphasizing different elements, it is imperative that managers fully understand their needs and desired outcomes on a project-level basis. When managers are faced with many requirements, responsibilities, and other external pressures, they require a method with a high level of efficacy that incorporates decision- making tools and adaptive management as a sustained active-learning process

(Possingham et al., 2001; Williams et al., 2007). I assessed two dominant adaptive management schools of thought in the literature to determine which approach is applied most successfully based on my a priori set of criteria. I related the success of each case study described in the literature to their assigned adaptive management approach. My goal was to increase efficacy of adaptive management approaches for natural-resources management by investigating the correlations among process, success, and efficacy of each approach.

In progressing from theoretical work in adaptive management to a practical application, I developed a population model for Least Terns and Piping Plovers on the

Lower Platte River (LPR), Nebraska. Although the LPR is relatively unaltered (Bental,

1982), there are multiple anthropogenic uses of the river that may impose external stressors on river functions and processes. These external stressors may adversely affect species for which the river provides habitat, particularly the endangered Interior Least

Tern (Sternula antillarum athalassos) and threatened Piping Plover (Charadius melodus;

U.S. Fish and Wildlife Service, 1985a; U.S. Fish and Wildlife Service, 1985b).

Therefore, the LPR is an ideal case study for developing models within an adaptive management context.

When adaptively managing populations, population projection models are used to evaluate management consequences based on current data (Nicolson et al., 2002;

Possingham et al., 2001; Williams et al., 2007). In an iterative decision-making context, projection models are continuously updated as more monitoring data becomes available which will influence future management decisions (Walters, 1986, 1997; Carpenter,

2002). I initiated model development at a rapid prototyping workshop organized and facilitated by Dr. Andrew J. Tyre and myself where invited participants specialized in the recovery of Least Terns and Piping Plovers on the LPR. The initial workshop goals were to define the management problem, objectives, and actions using structured decision- making. I developed the following decision problem to guide model development: How do alterations in available habitat affect the projected population sizes for the Least Tern and Piping Plover on the LPR, Nebraska? As the group developed multiple management objectives and an extensive list of competing actions, it became apparent that decision- making might benefit from developing a population model. To develop the population model, I estimated age-specific parameters for each species and the effect of habitat

factors on life history parameters from available monitoring data, expert opinion, and literature.

In contrast to the unaltered nature of the LPR, the central Platte River (CPR), once with wide and shallow braided channel morphology, now has stabilized banks, deeper channels, and increased flow (Kenney, 2000; Zallen, 1997; Echeverria, 2001).

Alterations to riverine habitat over the last several decades resulted from water diversions, land-use changes, and other basin alterations and contributed to the listing of various species (Sinokrot and Gulliver, 2000; Kroeger and McMurray, 2008; Smith,

2011). However, in contrast to the management efforts on the LPR, portions of the CPR, one of the largest Great Plains riverine ecosystems, are managed through an adaptive management based process under the Platte River Recovery Implementation Program

(hereafter simply Program) for four federally endangered and threatened species

(Schneider et al., 2005; Platte River Recovery Implementation Program, 2006).

To aid in the progression on the adaptive management plan, I offered a method of simplifying hypotheses and prioritizing research and management needs by providing results from simulated data scenarios analyzed in a multi-model analysis framework. I consolidated the Program’s research hypotheses into sets of statistical models for simulated multi-model analysis using generalized linear models. Such statistical models are relevant in evaluating management consequences based on current data within a decision-making context (Nicolson et al., 2002; Possingham et al., 2001; Williams et al.,

2007). I tested the ability of the monitoring program to distinguish between the hypotheses using simulated data to compare responses of Least Tern population numbers

under different research hypotheses on the CPR.

Models, whether conceptual or statistical, are “an abstraction or simplification of a natural phenomenon, developed to predict a new phenomenon or to provide insight into existing ones,” (Smith and Smith, 2006, p. 13). Within my research, I developed two types of models. The population model in my third chapter is a model that projects new

6 phenomenon based on monitoring data. In contrast, the simulation analysis in chapter four is a model that projects ten years of various scenarios of plausible monitoring data and provides insight into existing phenomenon about the data. Further, both models are derived from various hypotheses or “universal proposition[s] that suggests an explanation for some observed ecological situation,” (Sinclair et al., 2006, p. 394). While there may be controversy regarding the true differences between a model and a hypothesis, it is essential to understand the connection of hypotheses to management objectives. As adaptive management is a tool that offers natural-resources management a scientific method for confronting uncertainty while proceeding with decisions, adaptive management stresses the importance of testable and falsifiable objectives (Theberge et al., 2006). Testable and falsifiable objectives are necessary for proceeding with experiments in an adaptive management context.

Generally, within the scientific method, an experiment arbitrates amongst competing hypotheses or models (Griffith, 2001). However, there are two additional attributes that are important for distinguishing between methods of arbitrating among hypotheses. The first attribute is the number of simultaneous experimental treatments or the number of different manipulations of the system under the study area in use. This

could range from one (i.e., an observational study of existing conditions) to many treatments (i.e., a laboratory study with positive and negative control treatments and a response variable). The second attribute is the amount of replication within a treatment or the number of places and times the treatment effect was observed, which may also range from one replication to many.

In adaptive management, the purpose of an experiment is to use the management action itself to develop the experimental treatments. An experiment is more likely to determine the distinction between competing management actions faster than applying treatments sequentially to a single object as the simultaneity of the treatments in the experiment helps reduce the number of alternative explanations. Simultaneous alternative treatments are rarely conducted for large-scale manipulations of ecosystems, and it may not be helpful to describe such manipulations as experiments. This would lead to everything classified as an experiment and decrease the value of the word.

Though an experiment with simultaneous treatments is not the only method to distinguish between competing hypotheses (i.e., management actions), it is an excellent approach when possible.

I presented my thesis in five chapters with the first chapter introducing the need for my research and an overview of my thesis and relevant terminology. My second chapter is a review of adaptive management literature and assesses the level of success for two main schools of thought in adaptive management yielding useful recommendations for increased success in the implementation of adaptive management plans. In chapter three, I present a population model description focusing on the Interior

Least Tern and Piping Plover on the LPR, Nebraska, and how the model is a vital aspect of an adaptive management plan. In the fourth chapter, I report findings from a simulated multi-model analysis that assessed competing research hypotheses of an adaptive management plan by analyzing five simulated data scenarios with a multi-model

8 inference framework. The final chapter provides a synthetic summary of my conclusions from all chapters.

LITERATURE CITED

Bental, R., 1982. Nebraska’s Platte River, graphical analysis of flows. Conservation and

Survey Division, University of Nebraska, Lincoln, Nebraska, Nebraska Water

Survey Pap. 53. 47 p.

Berg, J., B. Bradshaw, J. Carbone, C. Chojnacky, S. Conroy, D. Cleaves, R. Solomon, S.

Yonts-Shepherd, 1999. Decision Protocol 2.0. U.S. Forest Service FS-634.

Berkes, F., 2004. Rethinking community-based conservation. Conservation Biology 18,

621–630.

Carpenter, S.R., 2002. Building an ecology of the long now. Ecology 83: 2069–2083.

Conroy, M.J., R.J. Barker, P.W. Dillingham, D. Fletcher, A.M. Gormley, I.M.

Westbrooke, 2008. Application of decision theory to conservation management:

recovery of Hector’s dolphin. Wildlife Research 35, 93–102.

Echeverria, J.D., 2001. No success like failure: the Platte River collaborative watershed

planning process. Wm. & Mary Envtl. L. & Pol’y Rev. 25, 559–603.

Gerber, L.R., J. Wielgus, E. Sala, 2007. A decision framework for the adaptive

management of an exploited species with implications for marine reserves.

Conservation Biology 21, 1594–1602.

Gregory, R., G. Long, 2009. Using structured decision making to help implement a

precautionary approach to endangered species management. Risk Analysis 29,

518–532.

Gregory, R.S., R.L. Keeney, 2002. Making Smarter Environmental Management

Decisions. Journal of the American Water Resources Association 38, 1601–1612.

Griffith, W.T., 2001. The Physics of Everyday Phenomena: A Conceptual Introduction to

Physics. McGraw-Hill Higher Education, New York, USA.

Gunderson, L.H., C.S. Holling, (Eds.), 2002. Panarchy: understanding transformations in

human and natural systems. Island Press, Washington, D.C.

Hilborn, R., C.J. Walters, 1981. Pitfalls of Environmental Baseline and Process Studies.

EIA Review 2, 265–278.

Holling, C. S., 2001. Understanding the complexity of economic, ecological, and social

systems. Ecosystems 4, 390–405.

Hughes, T.P., L.H. Gunderson, C. Folke, A.H. Baird, D. Bellwood, F. Berkes, B. Crona,

A. Helfgott, H. Leslie, J. Norberg, M. Nystrom, P. Olsson, H. Osterblom, M.

Scheffer, H. Schuttenberg, R.S. Steneck, M. Tengo, M. Troell, B. Walker, J.

Wilson, B. Worm, 2007. Adaptive Management of the Great Barrier Reef and the

Grand Canyon World Heritage Area. Ambio 36, 586–592.

Johnson, F.A., 2006. Adaptive harvest management and double-loop learning.

Transactions of the North American Wildlife and Natural Resources Conference

71, 197–213.

Johnson, F A., W.L. Kendall, J.A. Dubovsky, 2002. Conditions and limitations on 10

learning in the adaptive management of mallard harvests. Wildlife Society

Bulletin 30, 176–185.

Keeney, R.L., 1982. Decision analysis: an overview. Operations Research 30, 803–838.

Kenney, D.S., 2000. Arguing about consensus: Examining the case against western

watershed initiatives and other collaborative groups active in natural resources

management. Boulder, CO: Natural Resources Law Center, University of

Colorado School of Law.

Kroeger, T., A. McMurray, 2008. Economic benefits of conserving natural lands: case

study: central Platte biologically unique landscape, Nebraska. Prepared for the

Doris Duke Charitable Foundation. Washington, DC: Defenders of Wildlife. 46 p.

Lee, D.C., G.A. Bradshaw, 1998. Making monitoring work for managers. U.S. Forest

Service. Available from http://www.icbemp.gov/spatial/lee_monitor/begin.html

(accessed March 2010).

Levin, S. A., 1999. Fragile dominion: complexity and the commons. Perseus, New York.

Nicolson, C.R., A.M. Starfield, G.P. Kofinas, J.A. Kruse, 2002. Ten heuristics for

interdisciplinary modeling projects. Ecosystems 5, 376–384.

Platte River Recovery Implementation Program, 2006. Adaptive Management Plan.

Headwaters Corporation, Kearney, Nebraska. Available at

http://www.platteriverprogram.org (assessed November 2010).

Possingham, H.P., S.J. Andelman, B.R. Noon, S. Trombulak, H.R. Pulliam, 2001.

Making Smart Conservation Decisions, in: Orians, G., Soule, M., (Eds.), Research

Priorities for Conservation Biology. Island Press, Washington, DC, 1–18. 11

Reynolds, K., P. Cunningham, L. Bednar, M. Sanders, M. Foster, R. Olson, D. Schmolt,

D. Latham, B. Miller, J. Steffenson, 1996. A knowledge-based information

management system for watershed analysis in the Pacific Northwest U.S. AI

Applications 2, 9–22.

Robertson, D.P., R.B. Hull, 2001. Beyond biology: toward a more public ecology for

conservation. Conservation Biology 15, 970–979.

Ruitenbeek, J., C. Cartier, 2001. The invisible wand: adaptive co-management as an

emergent strategy in complex bio-economic systems. Occasional paper 34. Center

for International Forestry Research, Bogor, Indonesia, 47 p. Available from

http://www.cifor.cgiar.org (accessed March 2010).

Runge, M.C., F.A. Johnson, M.G. Anderson, M.D. Koneff, E.T. Reed, S.E. Mott, 2006.

The need for coherence between waterfowl harvest and habitat management. The

Wildlife Society Bulletin 34, 1231–1237.

Schneider, R., M. Humpert, K. Stoner, G. Steinauer, 2005. The Nebraska Natural Legacy

Project: A Comprehensive Wildlife Conservation Strategy. The Nebraska Game

and Parks Commission. Lincoln, Nebraska. August, 2005. 167 p.

Sinclair, A.R.E., J.M. Fryxell, G. Caughley, 2006. Wildlife ecology, conservation, and

management. 2nd edition. Blackwell Publishing, Malden, MA, USA.

Sinokrot, B.A., J.S. Gulliver, 2000. In-stream flow impact on river water temperatures.

Journal of Hydraulic Research, 38, 339–349.

Smith, C.B., 2011. Adaptive management on the central Platte River – Science, 12

engineering, and decision analysis to assist in the recovery of four species. Journal

of Environmental Management 92, 1414–1419.

Smith, T.M., R.L. Smith, 2006. Elements of Ecology. 6th Edition. Pearson Education,

Inc., San Francisco, CA, USA.

Theberge, J.B., M.T. Theberge, J.A. Vucetich, P.C. Paquet, 2006. Pitfalls of applying

adaptive management to a wolf population in Algonquin Provincial Park, Ontario.

Environmental Management 37, 451–460.

U.S. Fish and Wildlife Service (A). 1985. Endangered and threatened wildlife and plants:

Interior population of the least tern determination to be endangered. Federal

U.S. Fish and Wildlife Service (B). 1985. Determination of endangered and threatened

status for the piping plover. Federal Register 50, 50720–50734.

Walters, C., 1997. Challenges in adaptive management of riparian and coastal

ecosystems. Conservation Ecology 1, 1. Available from

http://www.consecol.org/vol1/iss2/art1 (accessed March 2010).

Walters, C. J., C.S. Holling, 1990. Large-scale management experiments and learning by

doing. Ecology 71, 2060–2068.

Walters, C.J., 1986. Adaptive management of renewable resources. MacMillan, New

York, USA.

Williams, B. K., R.C. Szaro, C.D. Shapiro, 2007. Adaptive Management: The U.S.

Department of the Interior Technical Guide. Adaptive Management Working 13

Group, U.S. Department of the Interior, Washington, D.C.

Williams, B.K., 1996. Adaptive optimization and the harvest of biological populations.

Mathematical Biosciences 136, 1–20.

Zallen, M., 1997. Integrating New Values with Old Uses in the Relicensing of Kingsley

Dam and Related Facilities: Making Part of the Problem a Part of the Solution.

Presented at Dams, Water and Power in the New West, Eighteenth Annual

Summer Conference of the Natural Resources Law Center, School of Law

University of Colorado, Boulder.

Chapter 2: EVALUATING THE EFFICACY OF ADAPTIVE MANAGEMENT

APPROACHES: IS THERE A FORMULA FOR SUCCESS?1

Abstract:

Within the field of natural-resources management, the application of adaptive management is appropriate for complex problems high in uncertainty. Adaptive management is becoming an increasingly popular management-decision tool within the scientific community and has developed into two primary schools of thought: the

Resilience-Experimentalist School (with high emphasis on stakeholder involvement, resilience, and highly complex models) and the Decision-Theoretic School (which results in relatively simple models through emphasizing stakeholder involvement for identifying management objectives). Because of these differences, adaptive management plans implemented under each of these schools may yield varying levels of success. I evaluated peer-reviewed literature focused on incorporation of adaptive management to identify components of successful adaptive management plans. My evaluation included adaptive management elements such as stakeholder involvement, definitions of management objectives and actions, use and complexity of predictive models, and the

sequence in which these elements were applied. I also defined a scale of degrees of success to make comparisons between the two adaptive management schools of thought.

My results include the relationship between the adaptive management process documented in the reviewed literature and my defined continuum of successful outcomes.

1 Published as: McFadden, J.E., T.L. Hiller, A.J. Tyre. 2011. Evaluating the efficacy of adaptive management approaches: Is there a formula for success? Journal of Environmental Management 92, 1354– 1359.

My data suggests an increase in the number of published articles with substantive discussion of adaptive management from 2000 to 2009 at a mean rate of annual change of

0.92 (r2 = 0.56). Additionally, my examination of data for temporal patterns related to each school resulted in an increase in acknowledgement of the Decision-Theoretic School of thought at a mean annual rate of change of 0.02 (r2 = 0.6679) and a stable acknowledgement for the Resilience-Experimentalist School of thought (r2 = 0.0042; slope = 0.0013). Identifying the elements of successful adaptive management will be advantageous to natural-resources managers considering adaptive management as a decision tool.

INTRODUCTION

Natural-resources managers are faced with value-laden decisions high in complexity, risk, and uncertainty (Levin, 1999; Gunderson and Holling, 2002; Berkes,

2004). The application of conventional research methods is often insufficient to support effective decision-making under these circumstances, particularly when decisions must be made regardless of the level of knowledge or uncertainty. There is a critical need to improve how research is incorporated into management decisions where uncertainty places limitations on contributions of science (Reynolds et al., 1996; Lee and Bradshaw,

1998; Berg et al., 1999; Robertson and Hull, 2001). Complex decisions involving risk in business and economics are often approached using structured decision-making (SDM), described by proponents as "a formalization of common sense for situations too complicated for the informal use of common sense" (Keeney, 1982, p. 806). Although such formal decision-making skills may be underdeveloped by natural-resources managers, the use of SDM is becoming more prevalent within the field of natural resources (Gregory and Keeney, 2002; Conroy et al., 2008; Gregory and Long, 2009).

By repeating decisions within an SDM approach, natural-resources managers can learn

through an ongoing process of implementing various management actions, monitoring 17

management outcomes, and updating ecological models by comparing actual outcomes with expected outcomes (Hilborn and Walters, 1981; Walters, 1986; Williams, 1996;

Carpenter, 2002; Johnson et al., 2002). This ongoing (i.e., iterative) learning process, learning by doing, is known as adaptive management.

Holling, 1990; Walters, 1997). Though there appears to be reluctance by some natural- resources managers to use adaptive management (Blumstein, 2007), it is becoming an increasingly popular concept and has developed within several governmental agencies, resulting in varying definitions of the process. However, there are commonalities amongst the various agencies regarding the adaptive management process, including establishing an iterative process that involves sharing of responsibilities and decision- making among managers, biologists, and stakeholders (Ruitenbeek and Cartier, 2001).

2007). However, adaptive management may result in variable degrees of success

(Walters, 1997). For natural-resources managers, it is important to improve

understanding of the adaptive management process by identifying correlates of success 18

within the available adaptive management literature. Managers can also apply learning by doing to the particular adaptive management approaches that have been implemented

(Johnson, 2006; Runge et. al, 2006; Williams et. al, 2007). In other words, it is necessary to “adaptively manage” the field of adaptive management by testing different decision-

19 making and modeling approaches, monitoring these management outcomes, and changing our practices to deliver better management outcomes.

Overall, there are two dominant adaptive management schools of thought

(Figure 1) which most adaptive management plans and approaches seem to follow: the

Resilience-Experimentalist Adaptive Management School which originates from the work of Gunderson et al. (1995), and the Decision-Theoretic Adaptive Management

School exemplified by Possingham et al. (2001) and the U.S. Department of Interior

(Williams et al., 2007). Management of the Florida Everglades (Walters et al., 1992;

Milon et al., 1997; Gunderson, 2000; Gunderson and Light, 2006) and the Glen Canyon

Dam Adaptive Management Program (Walters and Holling, 1990; Lee, 1999; Meretsky et al., 2000; Pulwarty and Melis, 2001) are well-known examples of the implementation of the Resilience-Experimentalist Adaptive Management School. In this school, there is high emphasis placed on obtaining a shared understanding among stakeholders of the system during the entire process, especially before defining objectives and management actions. In addition, proponents of this school also require active learning about ecosystem resilience, i.e., the capacity of an ecosystem to remain within its current state

or return to its original state following perturbation (Walters, 1997; Holling, 2001; 19

Hughes et al., 2007).

Alternatively, the Decision-Theoretic School, more heavily influenced by decision theory, also stresses communication among stakeholders, but communication is focused on defining the management problem, objectives, and actions prior to developing process models. The North American Waterfowl Management Plan (Johnson et al.,

1993; Johnson and Williams, 1999; Nichols et al., 2007) and conservation of red knots

(Calidris canutus) and horseshoe crabs (Limulus polyphemus) in Delaware Bay

(McGowan et al., 2009) are examples of the Decision-Theoretic School where decision theory approaches have been incorporated into the adaptive management process. These differences may appear minor, but the process for the Decision-Theoretic School often leads to less complex ecological models that are centered on the decision problem (e.g.,

Conroy et al., 2008), whereas the process for the Resilience-Experimentalist School leads to complex ecological models that include all potentially significant details of the ecosystem (e.g., Davis et al., 1994; Light and Dineen, 1994).

There are many organizations that promote adaptive management in ways that are broadly consistent with each school of thought (Table 1). The Adaptive Environmental

Assessment and Management process (Holling, 1978), Collaborative Adaptive

Management Network (2004), Sustainable Ecosystems Institute (2007), and Foundations of Success (2009) generally appear to follow the Resilience-Experimentalist School. The process of Adaptive Environmental Assessment and Management focuses on understanding dynamic environmental systems through developing computer simulations

under multiple management actions (Holling, 1978; Gunderson et al., 1995; Blann and 20

Light, 2000). Similarly, the involvement of the Collaborative Adaptive Management

Network in the management process is to facilitate adaptive management decisions, promote integrity and improved learning through collaboration of expertise, and serve as a primary role in adaptive management training of skilled managers in the field. The application of these various aspects of the Collaborative Adaptive Management Network

21 result in an increase in learning and efficacy of management plans. The Sustainable

Ecosystems Institute and Foundations of Success have similar roles in working with natural-resources agencies to develop adaptive management-based tools and decision- making strategies for providing natural-resources managers with problem-specific related facilitation, advising, and training services for individuals and organizations in need.

Figure 1. A comparison of the two dominant adaptive management schools of thought: the Resilience-Experimentalist Adaptive Management School and the Decision-Theoretic 21 Adaptive Management School.

Table 1. Comparison of five selected decision-making methods within the adaptive management literature including Gunderson’s et al. (1995) Adaptive Environmental Assessment and Management (AEAM), Possingham’s (2000) Structured Decision Making (SDM), Collaborative Adaptive Management Network (CAMNet; 2004), Department of Interior (DOI) Adaptive Management (AM) Protocol from the DOI AM Technical Guide (Williams et al., 2007), and Foundations of Success (FOS) with the Sustainable Ecosystem’s Institute (SEI, 2007). Comparison criteria include nine adaptive management related variables found from adaptive management literature along where variables were ordered (i.e. Order of Variables) according to their sequence within each decision-making method.

Adaptive Management Decision-Making Methods

Gunderson's et al. Possingham’s CAMNet DOI AM Protocol FOS & SEI Variable (1995) AEAM (2000) SDM (2004) (2007) (2007) 1. Stakeholder Yes; for Yes; entire Yes; entire Involvement Yes; entire process Yes; for objectives objectives process process Emphasis Yes; Key 2. Define Objectives Yes Yes Yes Yes Decision Points

3. Multiple Actions Yes Yes Yes Yes Yes

Yes; multiple Conceptual Yes; decision- 4. Predict Consequences competing hypothesis Modeling; rarely Yes No making protocol and modeling predictive Yes; specifically 5. Specify Constraints Yes Yes Yes Yes policy 6. Acknowledge Yes Yes Yes Yes No Uncertainty

Table 1. continued. Adaptive Management Decision-Making Methods

Gunderson's et al. Possingham’s CAMNet DOI AM Protocol FOS & SEI Variable (1995) AEAM (2000) SDM (2004) (2007) (2007) 7. Explicit Yes No Yes Yes Yes Experimentation

8. Monitoring Yes Yes Yes Yes Yes

9. Active Learning Yes No Yes Yes No Emphasis

Order of variables 1,4,5,2,3,6,7,8,9 2,3,5,6,4 1,2,4,6,7,8 2,3,4,5,6,7,8 1,2,3,5,7,8

The seven steps of SDM developed by Possingham et al. (2001) and Williams et al. (2007) are examples of the Decision-Theoretic school of thought. Possingham et al.

(2001) included monitoring and analysis of implemented management actions within the

SDM process. Here, the SDM process is designed to aid managers by developing ecological models to predict which action is best within the set of actions available.

Under this approach, natural-resources managers are provided with a method for prioritization of objectives and actions based on consequences of decisions and tradeoffs among objectives and active learning is achieved by requiring ongoing testing and re- evaluation of previous decisions. Similarly, in developing an Adaptive Management

Technical Guide and problem-scoping key, the Department of Interior provides aid for identification of appropriate problems and implementation of adaptive management

(Williams et al., 2007).

Despite the differences between schools discussed above, a recurrent theme in all adaptive management approaches is the ongoing monitoring of measurable objectives while also implementing selected actions (Walters and Holling, 1990; Field et al., 2004;

Gerber et al., 2005). With active learning and continuous monitoring, uncertainty

decreases and forecast management outcomes are more easily predicted (Walters, 1986, 24

1997). This allows for more informed decision making as the number of iterations increase in the adaptive management process.

While promotion of individual approaches to adaptive management occurs, there is no comparative overview of different adaptive management approaches (i.e., schools of thought). Scientific literature acknowledges that for the successful application of

25 adaptive management, there must be a cumulative experience of the process through building a thorough understanding of the various elements (Gerber et al., 2007). Overall, with multiple approaches emphasizing different elements, it is imperative that managers fully understand their needs and desired outcomes on a project-level basis. When managers are faced with many requirements, responsibilities, and other external pressures, they require a method with a high level of efficacy that incorporates decision- making tools and adaptive management as a sustained active-learning process. My objective was to assess the two dominant adaptive management schools of thought in the literature to determine which approach is applied most successfully based on my a priori set of criteria. I related the success of each case study described in the literature to their assigned adaptive management approach (i.e., Decision-Theoretic, Resilience-

Experimentalist, Other). My goal was to increase efficacy of adaptive management approaches for natural-resources management by investigating the correlations among process, success, and efficacy of each approach.

METHODS

I searched a selection of peer-reviewed literature for published case studies 25

incorporating adaptive management approaches to evaluate how successful outcomes vary by adaptive management school of thought. I selected eight scientific journals in the top ranks of ecology, conservation biology, and fisheries and wildlife management. I searched all articles from 2000 to 2009, unless limited to a shorter period by access, within The Journal of Wildlife Management, Ecology, Conservation Biology,

Conservation Ecology (2000–2003), Ecological Applications, Journal of Applied

Ecology, Wildlife Research (2008–2009), and Canadian Journal of Fisheries and Aquatic

Sciences. In selecting case studies for review, I required an article to contain the term

“adaptive management” within the document text. For my analysis of all adaptive management articles, I used a linear regression to describe the relationship of adaptive management articles as a function of time and the coefficient of determination (r2) to quantify the model fit.

To evaluate the success of different adaptive management schools of thought, I first defined success. The Merriam-Webster dictionary defines success as “a favorable or desired outcome” (Merriam-Webster, 2010). In applying this definition to the adaptive management process, there can be a wide range of outcomes considered successful. For example, Plan A may be more successful than Plan B if Plan A engaged in more active learning through implementing management actions over several years. Alternatively,

Plan A may be less successful if only Plan B met its specified objectives. Formal analysis of a decision problem, meeting objectives, engaging in active learning, and implementing management actions are all vital steps during the adaptive management

process. In arriving at a definition for success, I asked four questions: Was an explicit 26

formal analysis of the decision conducted? Does the resulting management plan include an iterative cycle? Was a management action implemented? Did the implemented action achieve the desired outcome? For purposes of this analysis, I acknowledged that there is a range of “successful” adaptive management up to and including achieving objectives and implementing actions from which management can learn.

I described five hierarchical categories (Mention, Theory, Suggest, Framework, and Implement) and divided articles according to the extent to which adaptive management was implemented based on information within each article. The Mention category included articles that used adaptive management merely as a catch phrase; these were not directly included in the analysis. The Theory category included articles discussing adaptive management in a general theoretical context about the application of adaptive management practices, but which lacked a description of a specific case study.

The Suggest category included articles acknowledging adaptive management as an appropriate approach for a particular management problem or management practice, but that did not provide a complete analysis of a specific problem. The Framework category described articles that, in addition to acknowledging adaptive management as an appropriate approach, provided a decision-based framework for a specific management problem. The Implement category described articles where a management action was implemented, the outcome monitored, and the results incorporated into the next management decision. This category also included articles where improvements were incorporated to an existing adaptive management framework. I assigned articles to the

category Against if they deemed adaptive management an inappropriate approach for 27

their management problem.

Case studies categorized as Framework or Implement articles were required to have stated objectives relevant to adaptive management, and have more than one management action to choose from for implementation. I established a list of variables found in the articles used for decision-making and management, including measurable

28 objectives, defined actions, stakeholder involvement, forecasted consequences, legal obligation, and action implementation. I defined these variables and the order in which they appeared throughout the adaptive management process for each case study. To compare case studies further, I identified the most appropriate school of thought for each based on original descriptions of each approach (e.g., Gunderson et al., 1995;

Possingham et al., 2001). Using the average number of case studies per success category,

I obtained the mean level of success for each approach. For my analysis of success categories, I used a linear regression to describe the relationship of the proportion of articles in each success category as a function of time and the coefficient of determination

(r2) to quantify the model fit. I used similar methods for my analysis of schools of thought where the proportion of articles in a school of thought is a function of time. To evaluate the relationship between success and a specific adaptive management school, I identified patterns of adaptive management variables within both schools that yielded similar levels of success.

RESULTS

I identified 96 scientific articles from eight scientific journals with some 28

substantive reference of the term adaptive management and found a basic temporal trend regarding discussion of adaptive management (Figure 2). My data showed an increase in number of published articles with substantive discussion of adaptive management from

2000 to 2009 at a mean rate of annual change of 0.92 (r2 = 0.5574), or about one article per year. Of my reviewed literature, I assigned 18% (n = 17) of articles to Theory, 42%

(n = 40) to Suggest, 24% (n = 23) to Framework, 14% (n = 13) to Implement, and 3% (n

= 3) to Against. The number of published articles that reported implementation of management actions within an adaptive management framework was low (24%) within my selected journals and years. In addition, I found three articles advising against the general concept of adaptive management, usually suggesting that adaptive management was not a practical approach for their particular study. For a complete list of my reviewed literature by school of thought and success category refer to Appendix A in the supplemental material.

I found unique trends for each success category, particularly for Theory and

Suggest categories, over time. For Theory, I observed a slight decrease in the proportion of articles discussing adaptive management in concept at a mean annual rate of change of

-0.02 over the last ten years (r2 = 0.1907), but found an increase in the proportion of articles in the Suggest category at a mean rate of annual change of 0.03 (r2=0.2442;

Figure 3). There was no conclusive trend in the percentage of the Framework (r2 =

0.0026; m = -0.003) and Implement (r2 = 0.0238; m = -0.0076) categories since 2000.

Figure 2. Number of scientific articles that reference the term adaptive management (n = 96; not including the Mention articles) from eight scientific journals by year from 2000 to 2009. Linear regression suggested a slope of 0.92 and r2 = 0.56.

Figure 3. The percentage of scientific articles in the suggest category over time from 2000–2009. Linear regression suggested a slope of m = 0.0346 and r2=0.2442.

After I sub-divided each category into the two schools of thought, I assigned 20%

(n = 9) of articles to Theory, 39% (n = 18) to Suggest, 30% (n = 14) to Framework, and

11% (n = 5) to Implement within the Resilience-Experimentalist School of Thought

(Figure 4). I assigned 0% (n = 2) of articles to Theory, 26% (n = 6) to Suggest, 35% (n =

8) to Framework, and 30% (n = 7) to Implement within the Decision-Theoretic School of

Thought. My examination of data for temporal patterns related to each school resulted in an increase in acknowledgement of the Decision-Theoretic School of thought at a mean annual rate of change of 0.02 (r2 = 0.6679) and a stable acknowledgement for the

Resilience-Experimentalist School of thought (r2 = 0.0042; m = 0.0013; Figure 5).

Figure 4. The percentage of scientific articles by success category for the Resilience- Experimentalist School of Thought (20% (n = 9) to Theory, 39% (n = 18) to Suggest, 30% (n = 14) to Framework, and 11% (n = 5) to Implement), the Decision-Theoretic School of Thought (0% (n = 2) to Theory, 26% (n = 6) to Suggest, 35% (n = 8) to Framework, and 30% (n = 7) to Implement), and other (25% (n = 6) to Theory, 67% (n = 16) to Suggest, 0% (n = 1) to Framework, and 0% (n = 1) to Implement).

Figure 5. The percentage of scientific articles with some reference of the term adaptive management (not including the Mention articles) from eight scientific journals by year from 2000 to 2009 categorized by two adaptive management schools, the Resilience- Experimentalist with r2 = 0.00; m = 0.0013; n = 10 (41 total articles) and the Decision- Theoretic with r2 = 0.67; m = 0.0232; n = 10 (24 total articles).

DISCUSSION

Based on my results, I have evidence that the amount of published literature

related to adaptive management has increased over the last decade, at least within the limited set of selected journals. In addition, the increase was not uniform among success 32

32 categories. I originally expected the Theory and Suggest articles to decrease and the

Framework and Implement articles to increase over time as an indication of increased acceptance and use of adaptive management. However, although Theory articles slightly decreased over time, the observed increase was in Suggest articles rather than Framework or Implement. It appears the current movement of adaptive management in practice is from discussion in a conceptual sense to a realization of the tool being useful in a

33 practical manner, but perhaps not yet to implementation. This suggests that the amount of time for theory to reach practice may be longer than the period of my analysis. While managers in the field of natural resources generally acknowledge adaptive management as an appropriate approach for managing complex ecosystems, the managers may experience difficulty in proceeding with the adaptive management process to the implementation stage. As suggested by Hobbs and Hilborn (2006), one difficulty in applying adaptive management in its original design by Holling (1978) lies in a lack of natural-resources researchers and managers trained in SDM, adaptive management, maximum likelihood, and Bayesian methods (Powell et. al, in press). Alternatively, it may be that successful implementations do not generate publishable articles, either because of a lack of interest on the part of managers in publishing, or because journal editors and referees do not regard such articles as worthy of publication.

The distribution of articles among categories differed for each school.

Numerically, the Resilience-Experimentalist School contained more Suggest and

Framework articles than the Decision-Theoretic School, but proportionally, the Decision-

Theoretic School had more Framework and Implement articles than Suggest articles. The

difference between the distribution of categories for each school may show that the 33

Decision-Theoretic School is easier to use for developing frameworks for natural- resources management.

It appears the Decision-Theoretic School provides a framework more conducive to implementing a management action than the Resilience-Experimentalist School, as there were proportionally more Implement articles under the Decision-Theoretic School.

The frameworks developed under the Decision-Theoretic School may result in higher efficacy because the Decision-Theoretic framework utilizes simple models to make decisions (Possingham et al., 2001). In turn, increased efficacy in the process may lead to an easier documentation process explaining the higher percentage of Framework and

Implement articles for the Decision-Theoretic School.

An equally important difference, experimentation, may also yield higher difficulty in management implementation for those following the Resilience-Experimentalist

School; in particular, the risk that an experiment will fail to achieve the management objective is a substantial barrier to achieving management implementation (Gregory et al., 2006). According to the Decision-Theoretic School, experiments are not required, but can be replaced with tradeoff analysis in situations where it is difficult to implement controlled experiments in large-scale ecosystems (McCarthy and Possingham, 2007).

While the exact mechanism causing such a difference between schools regarding number of Framework articles is unknown, recent case studies demonstrate multiple barriers to management implementation success. Such barriers include modeling difficulties, institutional rigidity, high financial costs, stakeholder dissention, and high political risks

(Hilborn and Walters, 1981; Walters, 1997; Gunderson, 1999; Sutherland, 2006). 34

My findings may be biased to some extent by my definitions of adaptive management and success. Given the vague linguistic nature of some literature reviewed for my study, the categorization of case studies is and must be subjective to some degree.

Additionally, I looked at relatively broad definitions of schools of thought because each approach may evolve by some unknown, but probably small, rate. I assumed that the

35 broad framework within each school did not evolve enough through time to affect my results, which covered a relatively short period (2000 to 2009).

Scientific literature acknowledges that successful application of adaptive management requires building a thorough understanding of the various elements of the process through cumulative experience (Gerber et al., 2007). My study takes the first meta-analytical perspective on adaptive management, explicitly recognizing and comparing different approaches and definitions of the process. Regardless of the challenge of publishing adaptive-management work that is applied in comparison to theoretical, I may see a longer delay in published works categorized as Framework and

Implement due to the time scale of implementing adaptive management given the slow transfer of technology. If adaptive management is to improve as an approach to management under uncertainty, it is imperative to study the process of adaptive management itself, including all approaches. My study evaluated two dominant schools of thought in the adaptive management field and showed that adaptive management as a concept continues to evolve through shifts in the dominant school of thought, as well as gain greater acceptance as a possible framework for management.

LITERATURE CITED

Berg, J., B. Bradshaw, J. Carbone, C. Chojnacky, S. Conroy, D. Cleaves, R. Solomon, S.

Yonts-Shepherd, 1999. Decision Protocol 2.0. U.S. Forest Service FS-634.

Berkes, F., 2004. Rethinking community-based conservation. Conservation Biology 18,

621–630.

Blann, K., S.S. Light, 2000. The path of last resort: adaptive environmental assessment

and management (AEAM). Adaptive Management Practitioners’ Network,

Minneapolis, Minnesota.

Blumstein, D.T., 2007. Darwinian decision making: putting the adaptive into adaptive

management. Conservation Biology 21, 552–553.

Carpenter, S.R., 2002. Building an ecology of the long now. Ecology 83: 2069–2083.

Collaborative Adaptive Management Network, 2004.

http://www.adaptivemanagement.net Accessed 26 February 2010.

Conroy, M.J., R.J. Barker, P.W. Dillingham, D. Fletcher, A.M. Gormley, I.M.

Westbrooke, 2008. Application of decision theory to conservation management:

recovery of Hector’s dolphin. Wildlife Research 35, 93–102.

Davis, S.M., L.H. Gunderson, W.A. Park, J.R. Richardson, J.E. Mattson, 1994. 36

Landscape dimension, composition, and function in a changing Everglades

ecosystem, in: Davis, S.M., J.C. Ogden, (Eds.), Everglades: the ecosystem and its

restoration. St. Lucie Press, Delray Beach, Florida, pp. 419–444.

Field, S. A., A.J. Tyre, N. Jonzen, J.R. Rhodes, H.P. Possingham, 2004. Minimizing the

cost of environmental management decisions by optimizing statistical thresholds.

Ecology Letters 7, 669–675.

Foundations of Success, 2009. Conceptualizing and Planning Conservation Projects and

Programs: A Training Manual. Foundations of Success, Bethesda, Maryland,

USA.

Gerber, L.R., J. Wielgus, E. Sala, 2007. A decision framework for the adaptive

management of an exploited species with implications for marine reserves.

Conservation Biology 21, 1594–1602.

Gerber, L. R., M. Beger, M.A. McCarthy, H.P. Possingham, 2005. A theory for optimal

monitoring of marine reserves. Ecology Letters 8, 829–837.

Gregory, R., G. Long, 2009. Using structured decision making to help implement a

precautionary approach to endangered species management. Risk Analysis 29,

518–532.

Gregory, R., D. Ohlson, J. Arvai, 2006. Deconstructing adaptive management: criteria for

applications to environmental management. Ecological Applications 16, 2411–

2425. 37

Gregory, R.S., R.L. Keeney, 2002. Making Smarter Environmental Management

Decisions. Journal of the American Water Resources Association 38, 1601–1612.

Gunderson, L, S.S. Light, 2006. Adaptive management and adaptive governance in the

Everglades ecosystem. Policy Sciences 39, 323–334.

Gunderson, L.H., C.S. Holling, (Eds.), 2002. Panarchy: understanding transformations in

human and natural systems. Island Press, Washington, D.C.

Gunderson, L.H., 2000. Ecological resilience – in theory and application. Annual Review

of Ecology and Systematics 31, 425–439.

Gunderson, L., 1999. Resilience, flexibility and adaptive management – antidotes for

spurious certitude? Conservation Ecology 3, 7.

http://www.consecol.org/vol3/iss1/art7/. Accessed 30 March 2010.

Gunderson, L., C. Holling, S. Light, 1995. Barriers and bridges for the renewal of

regional ecosystems. Columbia University Press, New York, New York, USA.

Hilborn, R., C.J. Walters, 1981. Pitfalls of Environmental Baseline and Process Studies.

EIA Review 2, 265–278.

Hobbs, N.T., R. Hilborn, 2006. Alternatives to statistical hypothesis testing in ecology: a

guide to self teaching. Ecological Applications 16, 5–19.

Holling, C. S., 2001. Understanding the complexity of economic, ecological, and social

systems. Ecosystems 4, 390–405.

Holling, C.S., (Ed.), 1978. Adaptive environmental assessment and management. Wiley,

London, UK. 38

Hughes, T.P., L.H. Gunderson, C. Folke, A.H. Baird, D. Bellwood, F. Berkes, B. Crona,

A. Helfgott, H. Leslie, J. Norberg, M. Nystrom, P. Olsson, H. Osterblom, M.

Scheffer, H. Schuttenberg, R.S. Steneck, M. Tengo, M. Troell, B. Walker, J.

Wilson, B. Worm, 2007. Adaptive Management of the Great Barrier Reef and the

Grand Canyon World Heritage Area. Ambio 36, 586–592.

Johnson, F.A., 2006. Adaptive harvest management and double-loop learning.

Transactions of the North American Wildlife and Natural Resources Conference

71, 197–213.

Johnson, F A., W.L. Kendall, J.A. Dubovsky, 2002. Conditions and limitations on

learning in the adaptive management of mallard harvests. Wildlife Society

Bulletin 30, 176–185.

Johnson, F.A., B.K. Williams, 1999. Protocol and practice in the adaptive management of

waterfowl harvests. Conservation Ecology 3, 9.

http://www.consecol.org/vol3/iss1/art9. Accessed 30 March 2010.

Johnson, F.A., B.K. Williams, J.D. Nichols, J.E. Hines, W.L. Kendall, G.W. Smith, D.F.

Caithamer, 1993. Developing an adaptive management strategy for harvesting

waterfowl in North America. Transactions of the North American Wildlife and

Natural Resources Conference 58, 565–583.

Keeney, R.L., 1982. Decision analysis: an overview. Operations Research 30, 803–838.

Lee, K.N., 1999. Appraising adaptive management. Conservation Ecology 3, 3.

http://www.consecol.org/vol3/iss2/art3/. Accessed 13 June 2010.

Lee, D.C., G.A. Bradshaw, 1998. Making monitoring work for managers. U.S. Forest 39

Service. http://www.icbemp.gov/spatial/lee_monitor/begin.html. Accessed 30

March 2010.

Levin, S. A., 1999. Fragile dominion: complexity and the commons. Perseus, New York.

Light, S.S., J.W. Dineen, 1994. Water control in the Everglades: a historical perspective

in: Davis, S.M., Ogden, J.C., (Eds.), Everglades: the ecosystem and its restoration.

St. Lucie Press, Delray Beach, Florida, pp. 47–84.

McCarthy, M.A., H.P. Possingham, 2007. Active adaptive management for conservation.

Conservation Biology 21, 956–963.

McGowan, C.P., D.R. Smith, J.D. Nichols, J. Martin, J.A. Sweka, J.E. Lyons, L.J. Niles,

K. Kalasz, R. Wong, J. Brust, M. Davis, 2009. A framework for the adaptive

management of Horseshoe Crab harvest in the Delaware Bay constrained by Red

Knot conservation. Stock Assessment Report No. 09-02 (Supplement B) of the

Atlantic State Marine Fisheries Commission.

http://www.asmfc.org/speciesDocuments/horseshoeCrab/annualreports/stockassm

treports/2009DelawareBayARMReport.pdf. Accessed 13 June 2010.

Meretsky, V.J., D.L. Wegner, L.E. Stevens, 2000. Balancing endangered species and

ecosystems: a case study of adaptive management in Grand Canyon.

Environmental Management 25, 579–586.

Merriam-Webster Online Dictionary, 2010. success. http://www.merriam-webster.com.

Accessed 26 February 2010. 40

Milon, J.W., C.F. Kiker, D.J. Lee, 1997. Ecosystem management and the Florida

Everglades: the role of social scientists. Journal of Agricultural and Applied

Economics 29, 99–107.

Nichols, J.D., M.C. Runge, F.A. Johnson, B.K. Williams, 2007. Adaptive harvest

management of North American waterfowl populations: a brief history and future

prospects. Journal of Ornithology 148, 343–349.

Possingham, H.P., S.J. Andelman, B.R. Noon, S. Trombulak, H.R. Pulliam, 2001.

Making Smart Conservation Decisions, in: Orians, G., Soule, M., (Eds.), Research

Priorities for Conservation Biology. Island Press, Washington, DC, 1–18.

Possingham, H.P., 2000. Is Environmental Research a Waste of Time? Ecological

Management & Restoration 1, 81–82.

Powell, L.A., A.J. Tyre, M.J. Conroy, J.T. Peterson, B.K. Williams, In Press. Integrating

adaptive management into wildlife curricula. Wildlife Professional.

Pulwarty, R.S., T.S. Melis, 2001. Climate extremes and adaptive management on the

Colorado River: lessons from the 1997-1998 ENSO event. Journal of

Environmental Management 63, 307–324.

Reynolds, K., P. Cunningham, L. Bednar, M. Sanders, M. Foster, R. Olson, D. Schmolt,

D. Latham, B. Miller, J. Steffenson, 1996. A knowledge-based information

management system for watershed analysis in the Pacific Northwest U.S. AI

Applications 2, 9–22. 41

Robertson, D.P., R.B. Hull, 2001. Beyond biology: toward a more public ecology for

conservation. Conservation Biology 15, 970–979.

Ruitenbeek, J., C. Cartier, 2001. The invisible wand: adaptive co-management as an

emergent strategy in complex bio-economic systems. Occasional paper 34. Center

for International Forestry Research, Bogor, Indonesia,

http://www.cifor.cgiar.org. 30 March 2010.

Runge, M.C., F.A. Johnson, M.G. Anderson, M.D. Koneff, E.T. Reed, S.E. Mott, 2006.

The need for coherence between waterfowl harvest and habitat management. The

Wildlife Society Bulletin 34, 1231–1237.

Sustainable Ecosystems Institute, 2007. http://sei.org/about.html. Accessed 26 February

2010.

Sutherland, W.J., 2006. Predicting the ecological consequences of environmental change:

a review of the methods. Journal of Applied Ecology 43, 599–616.

Walters, C., 1997. Challenges in adaptive management of riparian and coastal

ecosystems. Conservation Ecology 1, 1. http://www.consecol.org/vol1/iss2/art1.

Accessed 30 March 2010.

Walters, C., L. Gunderson, C.S. Holling, 1992. Experimental policies for water

management in the Everglades. Ecological Applications 2, 189–202.

Walters, C. J., C.S. Holling, 1990. Large-scale management experiments and learning by

doing. Ecology 71, 2060–2068.

Walters, C.J., 1986. Adaptive management of renewable resources. MacMillan, New 42

York, USA.

Williams, B. K., R.C. Szaro, C.D. Shapiro, 2007. Adaptive Management: The U.S.

Department of the Interior Technical Guide. Adaptive Management Working

Group, U.S. Department of the Interior, Washington, D.C.

Williams, B.K., 1996. Adaptive optimization and the harvest of biological populations.

Mathematical Biosciences 136, 1–20.

Chapter 3: PROJECTING POPULATION RESPONSES TO MANAGEMENT ON

THE LOWER PLATTE RIVER, NEBRASKA, FOR LEAST TERNS AND

PIPING PLOVERS WITH A QUANTITATIVE MODEL

Abstract:

In natural-resources management, adaptive management is appropriate for complicated problems with high levels of uncertainty. Adaptive management emphasizes stakeholder involvement, decision-making, and predictive models. The lower Platte

River (LPR) in Nebraska is a complicated ecosystem where resources management decisions affect endangered and threatened species such as the Interior Least Tern

(Sternula antillarum athalassos) and Piping Plover (Charadius melodus). Because there is high uncertainty associated with the responses of these two avian species to habitat restoration and other resource uses, a projection model for an adaptive management plan is valuable. I developed a projection population model for terns and plovers on the LPR modified from a similar existing model for the Missouri River ecosystem. My model estimates population sizes based on age-structured matrices for two areas of the LPR: on-

channel (i.e., sandbars) and off-channel (i.e., sandpits) breeding and nesting habitat. To 44

guide model development, I established two conceptual models for terns and plovers that visually showed the connection between processes and habitat and population state variables. I obtained all input parameters from experts, monitoring data from the Tern and Plover Conservation Partnership and the Nebraska Game and Parks Commission, and peer-reviewed literature. I included several sources of uncertainty including partial observability, structural uncertainty, and natural variation. I developed 13 management

45 scenarios that varied by the rate of habitat loss to examine the birds’ responses to changes in habitat availability. I conducted a sensitivity analysis to explore which input parameters had the greatest effect on population sizes for each avian species. Model results suggest that population sizes for each species respond similarly for short-term simulations, but differ for long-term scenarios. The sensitivity analysis suggested that hydro-peaking had the greatest effect on tern population size and productivity had the greatest effect on plover population sizes. The population model is a valuable tool in measuring the annual status of the two avian species and adaptively managing the habitat on the LPR. With multiple management objectives and competing management alternatives, this model will aid in testing implications of alternatives based on avian population’s responses to various habitat alterations. This projection model will be a useful technical tool for resources managers to select the best management alternative by comparing outcomes projected by the model for a set of alternatives. The model can be updated as monitoring of avian populations on the LPR continues. Further, the sensitivity analysis provides a method for prioritizing research needs. The ability of this quantitative model to adapt to new information makes it ideal for projecting management

implications for terns and plovers on the LPR within an adaptive management context. 45

INTRODUCTION

The lower Platte River (LPR), Nebraska, is a large and complicated river system that provides resources for a multitude of wildlife species in the Midwest. Given the relatively unaltered nature of the LPR (Bental, 1982), sandbars are more abundant on the

LPR than on the central Platte River (CPR) as the LPR retains characteristics of a braided channel and ephemeral sandbars (Kirsch, 1996; NRC, 2005; Parham, 2007). Broad channels with ephemeral sandbars are maintained on the LPR through periodic high flows with large amounts of sediment (Bental, 1982). However, there are multiple anthropogenic uses of river resources including mining operations, agriculture, and urban growth, which may impose external stressors on river functions and processes.

Additionally, the LPR provides habitat for several endangered and threatened species on the LPR, particularly the endangered Interior Least Tern (Sternula antillarum athalassos) and threatened Piping Plover (Charadius melodus).

The Least Tern was listed as a state and federally endangered species in 1985

(U.S. Fish and Wildlife Service, 1985a) and the Piping Plover as a state and federally threatened species in 1985 (U.S. Fish and Wildlife Service, 1985b). Although nesting

habitat provided along the LPR is no longer included in the critical habitat that was 46

designated in 2002 for the Northern Great Plains breeding populations, the birds do utilize the LPR for breeding and nesting. Various management agencies along the LPR pursue management actions focused on the recovery of each species.

Nesting habitat requirements for both species broadly includes bare or sparsely vegetated sand, such as river sandbars, shorelines, and off-river sandpits resulting from

47 mining operations (Ziewitz et al., 1992; Jenniges and Plettner, 2008). Monitoring by

Nebraska Game and Parks Commission suggests that sandpits are currently used for nesting habitat proportionally more than sandbars by birds of both species (Brown and

Jorgensen, 2009).

Brown and Jorgensen (2008) suggest the Least Tern population was relatively stable and the Piping Plover population was declining steadily on the LPR over the past two decades. As both species utilize bare sand at various locations (i.e., sandbars and sandpits) that potentially differ in resource availability, there may be differences in the productivity of the birds nesting at each site. Productivity differences may be aggravated by variations in sandbar quality throughout portions of the LPR, leading to possible changes in location of nesting activities (J. Jorgensen, Personal Communication).

However, as suggested by Palmer (2009), the quality of sandbars for nesting depends on river processes such as discharge and sediment load, and thus human activities.

Therefore, changes in nesting locations may be an indicator of decreased sandbar habitat quality (Schweitzer and Leslie, 1999; Joeckel and Henebry, 2008). In addition, sandpit sites along the LPR are desirable housing development sites (M. Brown, Personal

Communication), so under current river management, both habitat types are at risk of 47

declining, or continuing to decline.

Natural-resources managers of the LPR are confronted with a complicated ecosystem where the two avian species’ responses to habitat restoration and development are highly uncertain. One possible solution to decreasing the uncertainty about the species’ responses to habitat alterations is to implement management actions for the

48 recovery of endangered Least Tern and threatened Piping Plover in an adaptive management framework. Adaptive management is an iterative process that emphasizes stakeholder involvement, decision-making, uncertainty, and using models to formalize expected outcomes (Gregory and Keeney, 2002; Williams et al., 2007). Adaptive management is most effective for management decisions high in uncertainty where the decision-maker(s) has high controllability (Williams et al., 2007). When adaptively managing populations, population projection models are used to evaluate management consequences based on current data (Nicolson et al., 2002; Possingham et al., 2001;

Williams et al., 2007). In an iterative decision-making context, projection models are continuously updated as more monitoring data becomes available which will influence future management decisions (Walters, 1986, 1997; Carpenter, 2002).

I developed a simple population model for Least Terns and Piping Plovers on the

LPR that projects two state variables: habitat area and population size. I initiated model development at a rapid prototyping workshop organized and facilitated by Dr. Andrew J.

Tyre and myself. The workshop consisted of a small collaborative group of biologists and managers from the Nebraska Game and Parks Commission, the Tern and Plover

Conservation Partnership, and the University of Nebraska-Lincoln, specializing in the 48

recovery of Least Terns and Piping Plovers on the LPR. The initial workshop goals were to define the management problem, objectives, and actions using structured decision- making. I developed the following decision problem to guide model development: How do natural-resource managers best allocate resources, such as funding and time, for maintenance and restoration of the two habitat types, sandbars and sandpits, on the LPR

49 to achieve the recovery objectives of Least Terns and Piping Plovers over the next 20 years? As the group developed multiple management objectives and an extensive list of competing actions, it became apparent that decision-making might benefit from developing a population model. To develop the population model, I estimated age- specific parameters for each species and the effect of habitat factors on life history parameters from available monitoring data, expert opinion, and literature. This was used to determine the sensitivity of each population to each input parameter.

METHODS

STUDY AREA

The LPR consists of 103 river miles beginning at the confluence of the Loup

River and ending at the confluence of the Missouri River. Large sediment loads transported in intermittent high flows for sandbar maintenance occurs more often on the

LPR (Bental, 1982; Eschner, 1983; Rodekor and Engelbrecht, 1988; Kirsch, 1996; NRC,

2005) than in the CPR, Nebraska. Despite a more natural sediment load in the LPR, there

are factors imposing stress on the river that alter wildlife habitat and place restrictions on

management potential to restore or create riverine habitat. Such factors include urban 49

growth from Nebraska’s two largest cities, Lincoln and Omaha, mining operations with high water demands located throughout the LPR corridor, and a hydro-electric generating plant on the Loup River that traps sediment and alters the river’s hydrograph by introducing hydro-peaking. Although sandbar development is relatively sustained by

50 sedimentation in the river, alterations of wildlife habitat are likely to increase as a result from escalations in urban growth, mining, and hydro-power demands.

NESTING HABITAT TYPES

The LPR corridor contains two main habitat types for terns and plovers, sandbars and sandpits. Sandbars are unvegetated islands within the river channel, whereas sandpits are unvegetated areas of sand off-channel created by mining operations. As the birds require bare sand for nesting, they are able to utilize both habitat types. Kirsch

(1996) suggested that terns do not distinguish between habitats because they have similar, although highly variable, productivity in both habitats. Nonetheless, there may be a difference in the relative value for the birds. Brown and Jorgensen’s (2008, 2009) work suggested that tern nest and chick survival was greater for individuals nesting on sandbars than sandpits. Regardless of whether terns perceive differences between the two habitat types, it is important to determine how changes in the absolute and relative amounts of sandbars and sandpits affect the population dynamics of each species.

CONCEPTUAL MODEL

I developed a conceptual model of the population and habitat models that show

the processes that connect the state variables within the model (Figures 1 and 2). State 50

variables for the habitat model are the total area of both habitat types, and state variables for the population model are total number of individuals in each age class for each species. State variables of the population model are influenced by both habitat types from the habitat model through density and immigration and emigration from and to both habitat types.

Figure 1. Conceptual model depicting the interactions between the habitat model and Piping Plover population model. Bold square boxes display state variables and round- edged boxes show processes between state variables indicated through arrows. The habitat loss rates include changes in erosion and vegetation processes for sandbar area and include changes in development and mining operations that affect sandpit area.

Figure 2. Conceptual model depicting the interactions between the habitat model and Least Tern population model. Bold square boxes display state variables and round-edged boxes show processes between state variables indicated through arrows. The habitat loss rates include changes in erosion and vegetation processes for sandbar area and include changes in development and mining operations that affect sandpit area.

POPULATION MODEL

I developed a post-breeding age-structured population model (Caswell, 2001) that projects tern and plover populations through time in response to changes in habitat using

Microsoft Excel (Microsoft, 2007). I accounted for parameter uncertainty as well as demographic and environmental stochasticity within the system by utilizing Monte Carlo simulation. State variables for the population model were the number of individuals for each age class of each species in both habitats. The age classes are the number of plover after hatch year birds (AHY) and hatch year birds (HY) and total number of tern after second hatch year birds (ASY), HY birds, and second hatch year birds (SHY). The total population size consists of all breeding individuals for each species, AHYs and ASYs, from both habitat types and can be related to possible species recovery targets.

Additionally, hatch year ratio, a ratio of the number of HY to the number of AHY or

ASY birds, is calculated for both species as an output metric. As the model is based on monitoring data from annual post-breeding censuses, it calculates projections of population sizes and hatch year ratios annually based on the previous year. The short- term model currently simulates 10 years and the long-term model 100 years. The time

period of the model is flexible, and can be adapted to management requirements. 52

Population Dynamics

I established an age-structured matrix for both species where individuals are classified based on their differing reproductive and survival rates as functions of age

(Owen-Smith, 2007). This allows the model to calculate the changes in the number of

53 individuals over time of each age class at increments of 1 year. The age-structured matrix for plovers contains two survival rates (S) and two fecundity rates (F)

where S0 is the survival rate of HY birds and S1 is the survival rate of AHY birds. The tern age-structured matrix contains three survival rates and three fecundity rates

where S0 is the survival rate of HY birds, S1 is the survival rate of SHY birds, and S2 is the survival rate of ASY birds. As the ASY and AHY birds in each population have the same survival and reproduction rates from year to year, I represented all ASY and AHY age classes in one final age class according to the Usher matrix notation. I assumed effects of predation are accounted for in the model through the estimation of survival rates.

I incorporated dispersal between habitats into the model by assuming that individuals disperse before they are reproductively mature and continue to nest in the

same habitat type throughout adulthood from year to year. I accounted for demographic 53

53 stochasticity in fecundity and survival by using random variables with appropriate distributions. For the number of HY birds produced by ASY and AHY birds, I used a negative binomial distribution, which includes environmental stochasticity, demographic stochasticity, and demographic heterogeneity (Melbourne and Hastings, 2008). I used binomial distributions to calculate the number of ASY, AHY, and SHY survivors.

I used binomial distributions to incorporate demographic stochasticity in survival in the number of AHY birds:

where is the number of plover AHY birds for year t, is the survival rate for plover AHY birds in the current year, is the number of plover HY birds in the current year, is the number of plover emigrants in the current year, is the number of

plover immigrants in the current year, and is the survival rate for plover HY birds in the current year. I used the following equation to estimate the number of ASY terns with a binomial distribution:

where is the number of tern ASY birds for the following year, is the number

of tern ASY birds in the current year, is the survival rate for tern ASY birds in the current year, is the number of tern SY birds in the current year, is the number of plover emigrants in the current year, is the number of plover immigrants in the

current year, and is the survival rate for tern SY birds in the current year. I used the

following equation to estimate the number of SY birds for terns with a binomial

distribution:

where is the number of SY terns in the following year, is the number of HY

terns in the current year, and is the survival rate for HY terns in the current year. I

55 used the following equation to estimate the mean number of HY birds for both avian species:

where is the mean number of HY birds specific to the species ( , is the base of the natural log (2.718), is the bird productivity at , is the density at , is the habitat divisor to allow for scaling, is hydro-peaking effects at , is a switch for hydro- peaking where when = 1, hydro-peaking effects and when = 0, hydro-peaking does not affect . I used a negative binomial distribution to estimate both the number of

HY terns and HY plovers:

where is the number of HY birds at , is the negative binomial distribution, and is the carrying capacity. I used the following equations for each species to estimate the number of emigrants and immigrants:

where is the dispersal at and is the total area of available habitat at .

Both species are linked to the annual variation in habitat area through the ratio of habitat area to adult numbers. If the ratio is >0.01 acres / pair, the model will calculate an expected number of hatch year birds. Additionally, Kirsch (1996) suggested density- dependence in chick and egg mortality for terns and Catlin (2009) suggests density-

56 dependence in hatch year birds to recruitment for plovers. Given the findings of both studies, I incorporated the effects of density-dependence into the productivity as a function of nesting habitat area and population size. Therefore, as the density parameter has indirect negative effects on the population size through the projection of hatch year birds per year, the model simulates the relationship between population size and productivity where when adult numbers increase, productivity of hatch year birds decreases. I expected terns and plovers to respond differently to habitat modification and restoration depending on the strength of density-dependence exhibited by each species.

Although Kirsch (1996) found stronger effects of density-dependence on sandpits than sandbars, there was a great deal of variation in the data, and therefore I assumed similar effects of density-dependence for birds nesting and fledging on both habitat types for model simplicity.

Other sources of uncertainty incorporated into the model include partial observability, epistemic uncertainty, and natural variation. Partial observability is accounted for through detection rates, which assume that monitoring techniques do not detect each individual. I accounted for epistemic uncertainty using distributions placed

on input parameters, which allowed the input parameters to vary continuously between 56

runs. Natural variation over time is driven by and accounted for through variation in river discharge where specifically, environmental stochasticity affects productivity through habitat area, which in turn is affected by discharge. Additionally, effects from climate change will clearly alter discharge, but at this point climate change effects are unknown regarding discharge on the LPR. However, when climate change effects are

57 quantified, those effects can be incorporated into the model at that time. Until then, we assume that climate change will affect the birds through discharge.

Parameter Input

The population model contains three classes of input parameters, survival estimates, productivity components, and detectability, which are specific to each species

(Table 1). All input parameters were obtained from either monitoring data from

Nebraska Game and Parks Commission and the Tern and Plover Conservation

Partnership, current literature, or best estimates from experts. Estimates obtained from experts were assumed to represent the mean of the subjective belief about the probability distributions of each estimate. Normal distributions were placed on survival estimates and productivity components. I assumed that nest success and HY survival are equal for both habitat types (Kirsch, 2000).

Table 1. Input parameters and associated coefficient of variations and standard errors for tern and plover population model that were obtained from current literature, research and monitoring estimates, or best estimates from specialists. Input Temporal 57 Uncertainty Parameter Description Parameter 1 Variability Source 57

(CV ) Class (SE2) After Hatch Plover 0.74 6 0.0074 Larson et al., 2000 Survival Year Estimates Hatch Year 0.60 9 0.006 Expert3, 2008 Massey et al., 1992; Renken and Tern Survival After Second 0.80 6 0.0069 Smith, 1995; Estimates Year Thompson et al., 1997 Massey et al., Hatch Year 0.81 9 0.0081 1992

Table 1. Continued. Input Temporal Uncertainty Parameter Description Parameter Variability Source (CV1) Class (SE2) Tern Survival Second Massey et al., 0.81 9 0.0081 Estimates Hatch Year 1992 Bird USACE4 data, 0.38 47 0.019 Productivity 2004-2007 USACE data, Plover Density -0.22 31 0.0022 2004-2007 Productivity USACE data, Components Dispersion 0.64 22 0.0064 2004-2007 Hydro- Workshop -0.74 75 0.0075 peaking estimate, 2008 Bird USACE data, 0.58 47 0.0288 Productivity 2004-2007 USACE data, Tern Density -0.17 31 0.0016 2004-2007 Productivity USACE data, Components Dispersion 0.99 22 0.0099 2004-2007 Hydro- Workshop -0.74 75 0.0075 peaking estimate, 2008 After Workshop Tern and Second/ 0.70 estimate, 2008 Plover Hatch Year Detectability Workshop Hatch Year 0.95 estimate, 2008 1-4 CV = Coefficient of Variation (i.e., the amount of uncertainty placed on the parameter estimate), SE = Standard Error, Expert = biologist from United States Fish and Wildlife Service, USACE = United States Army Core of Engineers.

HABITAT MODEL

The hydrology of the river is a driving variable for processes on the river and 58

58 affects the productivity for birds nesting on sandbars (Licht, 2001). State variables for the habitat model include discharge, available habitat for sandbars and sandpits, and hydro-peaking. I obtained historical daily mean discharge in cubic feet per second (cfs) data from 04/01/1949 through 12/31/2009 from USGS Gauge No. 0679600 at North

Bend, Nebraska (Figure 3). I used the following equation to simulate discharge from historical data:

where is the simulated discharge (cfs) in the current year ( ) and is the historical discharge data (cfs) that allows the model to select a specific discharge associated with year . I assumed that available sandbar habitat is a function of discharge (e.g., during extremely high discharge there is little available sandbar habitat).

Parham (2007) found when discharge begins to increase, available large disconnected sandbar habitat increases exponentially until a threshold of around 5,480 cubic feet per second (cfs) when the habitat begins to decline. I adapted Parham’s (2007) findings, by scaling the predicted area based on the USGS Gauge data to estimated sandbar area for

2009 (Figure 4; J. Jorgensen, Personal Communication). I used the peak discharge from the daily mean discharges throughout the month of May as an approximation of the discharge birds would experience during nest initiation (Figure 5). I used the following equation as adapted from Parham’s (2007) work:

where is the percent of available habitat as a function of the best fit line for

based on historical data.

Figure 3. The historical mean daily discharge data of the lower Platte River from USGS Gauge No. 0679600 from 04/01/1949 through 12/31/2009 at North Bend, Nebraska.

Figure 4. The available sandbar habitat area on the lower Platte River as a function of discharge from Parham (2007) applied to the USGS Gauge No. 0679600 daily mean discharge data and habitat area data obtained from Nebraska Game and Parks Commission.

Figure 5. Selected peak mean daily discharge on the lower Platte River during the month of May from USGS Gauge No. 0679600 from 04/01/1949–12/31/2009 at North Bend, Nebraska.

For each simulation run, the model selected a random starting year between 1949 and 1999, and then used the actual observed sequence of annual discharge from the next

10 years of the historical discharge data. This procedure was modified for long-term runs

(model runs greater than 60 years), because the length of the simulation run exceeded the length of the available gauge data. In the long-term runs, the model randomly selected

the discharge variable from the entire dataset for each simulated year. While simple, this 61

procedure assumes all temporal correlations in discharge between years are zero, and so was only used to check the biological plausibility of the model over long time periods. I used the following equations to estimate available sandbar area:

where is the historical sandbar area estimated for the following year ( ), is

the habitat loss rate for sandbars, is the current sandbar area, and is the simulated area of sandbar habitat. I used the following equation to estimate the area of available sandpit habitat:

where is the area of simulated sandpit area for the following year, is the current sandpit area, and is the habitat loss rate for sandpits. I account for anthropogenic effects, such as development and mining operations, on the birds by assuming the birds are affected through habitat availability. Depending on the severity of human effects, can be altered appropriately.

Effects of hydro-peaking on the LPR documented by Elliot et al. (2009), present daily fluctuations in the area of available sandbar habitat. Hydro-peaking results from hydro-electric generating plants that require daily spikes in water discharge to meet supply demands. As terns and plovers require ephemeral and variable habitat (Kirsch,

1996; Thompson et al., 1997), hydro-electric plants aid in decreasing natural production

of sandbars by generating daily flow spikes that result in frequent sandbar inundation,

trap suspended sediment necessary for sandbar construction, and result in a stabilized 62

hydrograph. A stabilized hydrograph decreases extreme high flow events that support significant sandbar development. Therefore, hydro-electric plants pose threats to terns and plovers nesting downstream. With the Loup River Power District generating hydro- peaking events upstream from the LPR, I incorporated hydro-peaking as a parameter with negative effects on the production of HY birds.

SCENARIO DESCRIPTION

For demonstration purposes, I developed a set of thirteen scenarios that varied the rate of habitat loss for each type of habitat to assess the population consequences for terns and plovers (Table 2). I ran 1,000 iterations for each scenario over ten years, from 2008 through 2018. I selected two scenarios of the thirteen scenarios, the Status Quo and Total

Loss, to run out to 2100, simulating long-term habitat and population responses. Model runs initiate in 2008 as based on monitoring data during 2008 forecasting from the monitoring data and then projecting forward from 2008. There are multiple definitions of breeding and nesting habitat in the Great Plains region for terns and plovers (Faanes,

1983; Adolf, 2001; Marcus et al., 2007). Given the potential discrepancy in the definition of habitat (Hall et al., 1997; Hodges, 2008), it is important that I be clear in what I refer to as habitat in my study. The definition of habitat used to determine the area of available habitat incorporated in the model is any area of bare and open, unvegetated sand near a body of water and any mid-channel sandbar detached from the shoreline (M.

Brown, Personal Communication). However, the model is not dependent on a single definition of habitat, as long as the definition is kept in mind when comparing projections

with other results. The definition that will be most useful is dependent on the defined 63

objectives (Colyvan et al., 2009); it can be changed in the future if those objectives change. The scenarios were based on initial estimate of total habitat area in 2008 where sandbar area consisted of 218 hectares and sandpit area of 527 hectares as estimated by

Nebraska Game and Parks Commission and the Tern and Plover Conservation

Partnership.

The habitat loss rate for the scenarios is not intuitive. Given the Status Quo scenario with a habitat loss rate of 0 for each habitat type, there is no change in habitat area from one year to the next. However, the Sandpit Gain scenario with a habitat loss rate of -0.05 for sandpits results in a 5% increase in sandpit area with no change of sandbar area. Conversely, the Sandpit Loss scenario, with a habitat loss rate of 0.05 for sandpits, results in a 5% decrease in sandpit area annually. The habitat loss rates for all thirteen scenarios are meant to span the range of future possibilities. As 10% is a large change in habitat area, all scenarios do not necessarily reflect reality. However, 10% is a reasonable maximum for a habitat loss rate as it results in a 90% reduction of habitat over

20 years.

Table 2. Thirteen scenarios developed for model demonstration purposes only that vary by the loss rate of the specified habitat type area.

Variable

Scenario Description Sandbar Loss Sandpit Loss Status Quo (1) 0 0 Sandpit Gain (2) 0 -0.05

Sandpit Loss (3) 0 0.05

Extreme Sandpit Loss (4) 0 0.1 64

Sandbar Gain (5) -0.05 0 64

Sandbar Loss (6) 0.05 0 Extreme Sandbar Loss (7) 0.1 0 Total Gain (8) -0.05 -0.05 Total Loss (9) 0.1 0.1 Sandbar Gain & Sandpit Loss (10) -0.05 0.05 Sandbar Loss & Sandpit Gain (11) 0.05 -0.05 Status Quo Long-term (12) 0 0 Total Loss Long-term (13) 0.1 0.1

SENSITIVITY ANALYSIS

I conducted a sensitivity analysis of population size as a function of the input parameters for each species. Each input parameter was scaled to a zero mean and variance of 1 to allow for more accurate comparisons. I log transformed the response variables (i.e., population size) and fitted each analysis to a generalized linear model using the following equation for terns:

where the 2 indicates that all main effects and 2nd order interactions were included, and the following equation for plovers:

RESULTS

SCENARIO RESULTS

The Status Quo long-term scenario, with a habitat loss rate of 0.0, suggests that the plover population will increase and stabilize at a median of approximately 2,500

pairs, while the tern population stabilizes around 3,000 individuals by 2100 (Figure 6A). 65

However, the Total Loss long-term scenario, with an extreme habitat loss rate of 0.1 for both habitat types, suggests that the plover population continues to increase to just under

500 pairs before declining towards zero while the tern population increases to about 800 individuals before declining towards zero (Figure 6B). The causes for the differing responses to habitat loss between the two avian species are unknown and may imply the two species have unique limiting resources. The uncertainty associated with the causes

66 of their differences in responses result from summarizing the differences between the two species within the productivity and density-dependence estimates. There are many plausible reasons for the differences between tern and plover population sizes, including life cycle and foraging differences, but the model is unable to detect the exact explanation for differences in responses to habitat changes.

Figure 6. Results of the long-term model scenarios show Least Tern (long dash line) and Piping Plover (short dash line) median population responses to differences in habitat loss rates (solid line) over 92 years. Habitat loss rates for Status Quo Long-term Scenario are 0.00 for both habitat types (A) and are 0.10 for both habitat types in the Total Loss Long- term Scenario (B).

For short-term scenarios, model results suggest that, while all scenarios with a loss rate >0.0 suggest a decrease in habitat area, with the exception of the Sandbar Loss

& Sandpit Gain Scenario (Figure 7B), the Total Loss and Extreme Sandpit Loss express the greatest decrease in habitat area (Figures 8C and 9B). The decline in habitat area for the Extreme Sandpit Loss scenario reflects the initial habitat areas for the two habitat types, as the initial area for Sandpits is greater than the area of Sandbars. If 10% of

Sandpit area is lost annually, but Sandbar area remains stable, total available habitat stabilizes at the initial Sandbar area.

Figure 7. Results for the short-term model scenarios show Least Tern (long dash line) and Piping Plover (short dash line) median population responses to various differences in available habitat area (solid line). Habitat loss rates for Sandbar Gain & Sandpit Loss Scenario are -0.05 for sandbars and 0.05 for sandpits (A), and are 0.05 for sandbars and - 0.05 for sandpits in the Sandbar Loss & Sandpit Gain Scenario (B).

Figure 8. Results for the short-term model scenarios showing Least Tern (long dash line) and Piping Plover (short dash line) median population responses to differences in habitat loss rates (solid line). Habitat loss rates for Status Quo Scenario are 0.00 for both habitat types (A), are -0.05 for both habitat types in the Total Gain Scenario (B), and are 0.10 for both habitat types in the Total Loss Scenario (C).

Figure 9. Results for the short-term model scenarios show Least Tern (long dash line) and Piping Plover (short dash line) median population responses to various decreases in habitat area (solid line). Habitat loss rates for Sandpit Loss Scenario are 0.00 for sandbars and 0.05 for sandpits (A), are 0.00 for sandbars and 0.10 for sandpits in the Extreme Sandpit Loss Scenario (B), are 0.05 for sandbars and 0.00 for sandpits in the Sandbar Loss Scenario (C), and are 0.10 for sandbars and 0.05 for sandpits in the Extreme Sandbar Loss Scenario (D).

The plover and tern populations appear to respond positively over time to all short-term scenarios (Figures 7–10), especially for scenarios with a sandpit loss rate <0.0

(i.e., a gain in habitat; Figures 10A, 10B, 11B). While it appears the tern population size is larger than the plover population size, the difference between the two population sizes is due to the survey methods used for each species. Terns are surveyed as individuals, where as plovers are surveyed as breeding pairs.

Model results suggest that the medians over all 1,000 runs for each of the short- term scenarios greatly fluctuate at the conclusion of the period for each scenario regarding the projected habitat area (Figure 11). However, the median appears to be constant for the tern and plover population sizes. I determined the proportion of runs that during the final year of each scenario showed a decline in habitat area and population size

(Figure 12). The results suggest that as habitat loss increases, the proportion of runs resulting in less habitat area in the final year increases. The proportion of population sizes that decreased by the end of each short-term scenario remain steady, fluctuating between 0.13 and 0.18 for plovers and 0.63 and 0.68 for terns. However, for the long- term scenarios, the population sizes differ greatly between the Status Quo long-term and the Total Loss long-term scenarios.

Figure 10. Results for the short-term model scenarios show Least Tern (long dash line) and Piping Plover (short dash line) median population responses to increases in habitat area (solid line). Habitat loss rates for Sandpit Gain Scenario are 0.00 for sandbars and - 0.05 for sandpits (A) and are -0.05 for sandbars in the Sandbar Gain Scenario (B).

Figure 11. Model results showing the median (horizontal line in center for each box), 25th and 75th percentiles (ends of boxes), 10th and 90th percentiles (vertical lines, and outliers (open circles) over all 1,000 runs for each scenario at the conclusion of the model time frame. Initial starting habitat area varies over all scenarios with an average of 725±5.49 hectares.

Figure 12. The proportion of runs during the final year of each scenario that show a decline in habitat area and population size. Shown are short-term scenarios (1-11), long- term scenarios (12 and 13), habitat area (black bars), plover population size (dark gray bars), and tern population size (light gray bars).

SENSITIVITY ANALYSIS

Given that all input parameters vary within the model due to parameter uncertainty and natural temporal variation, the relationship between the population size and survival rates has a great deal of noise (Figure 13). However, the model does suggest that as survival rates increase, the population size increases. As I examined the effects on

population size of selected input parameters, the analysis suggested that hydro-peaking

had the greatest effect on tern population size, followed by the survival rates for each age 73

class within the model (Table 3). For plover population size, my analysis suggested that bird productivity had the greatest effect, followed by AHY survival (Table 4).

Figure 13. Model sensitivity output for Least Tern population, where much variation between the population size (i.e., Log Number of Adults) and the survival rates are attributed to natural variation and that input parameters vary throughout the time frame of each model run. The population model suggests that population size increases as survival rates of after second year birds (A), fledglings (B), and second year birds (C) increase.

Table 3. Sensitivity analysis results for Least Tern data where the fitted linear model estimates show an overall p-value of <2.2e-16 with r2 = 0.955 and F = 1011. Covariate Estimate Std. Error t value (Intercept) 5.92 0.00 1747.12 Hydro-peaking 1.03 0.00 303.50 After Second Year Survival (ASY) 0.33 0.00 96.83 Hatch Year Survival (HY) 0.18 0.00 53.85 Second Year Survival (SY) 0.16 0.00 47.08 Bird Productivity (BP) 0.10 0.00 28.88 Density 0.06 0.00 17.39 Density * Hydro-peaking 0.04 0.00 10.93 HY * Hydro-peaking 0.03 0.00 9.88 SY * Hydro-peaking 0.03 0.00 8.28 BP * Hydro-peaking 0.02 0.00 5.51 SY * BP 0.01 0.00 1.56 HY * SY 0.00 0.00 1.22 ASY * Density 0.00 0.00 1.13 SY * Density 0.00 0.00 1.02 HY * BP 0.00 0.00 0.92 BP * Density 0.00 0.00 0.15 HY * Density 0.00 0.00 -0.40 ASY * SY 0.00 0.00 -1.17 ASY * HY -0.01 0.00 -2.77 ASY * BP -0.01 0.00 -3.55 ASY * Hydro-peaking -0.07 0.00 -19.56

Table 4. Sensitivity analysis for Piping Plover data where the fitted linear model 75

2 estimates show an overall p-value of <2.2e-16 with r = 0.786 and F = 240.9. 75

Covariate Estimate Std. Error t value (Intercept) 5.81 0.01 983.71 Bird Productivity (BP) 0.55 0.01 92.68 After Hatch Year Survival (AHY) 0.37 0.01 61.96 Hatch Year Survival (HY) 0.28 0.01 47.84 Hydro-peaking 0.27 0.01 45.64 Density 0.05 0.01 8.60 HY * BP 0.03 0.01 4.49 BP * Density 0.02 0.01 3.18

Table 4. Continued. Covariate Estimate Std. Error t value HY * Density 0.01 0.01 2.29 Density * Hydro-peaking 0.01 0.01 1.86 AHY * Density 0.01 0.01 1.68 HY * Hydro-peaking 0.01 0.01 1.36 BP * Hydro-peaking 0.00 0.01 0.41 AHY * HY -0.02 0.01 -3.51 AHY * Hydro-peaking -0.02 0.01 -3.66 AHY * BP -0.03 0.01 -5.88

DISCUSSION

When managing for endangered and threatened species, having the capability to project the annual population dynamics of the listed species is advantageous for resources managers confronted with numerous regulations (Rubin et al., 2002; Caswell and

Fujiwara, 2004). Natural-resource managers can utilize population models as a method of evaluation over a specified period where the model can provide quantitative estimations of existing management success and projections of the population’s future status. Alternatively, managers could estimate how an unforeseeable event could influence the system and decrease the possibility of surprises, to “expect the unexpected”

(Parma et al., 1998; Greeuw et al., 2000). Development of possible extreme events 76

through workshops with experts and stakeholders provides managers with a conceptual model of the effects of extreme events. By modifying the input parameters, such as the survival or habitat loss rates, to extreme values under unexpected events such as extreme high flows, the model output will aid in providing insight for resources managers on the inherent uncertainty of surprises.

To further decrease uncertainty within the model and increase the knowledge of the managed system, the sensitivity analysis conducted provided a method for guiding prioritization of future research projects. The sensitivity analysis suggested that the tern population size was most sensitive to effects of hydro-peaking on reproductive output, followed by the ASY and the SHY survival rates, while the plover population size was most sensitive to bird productivity, followed by AHY survival. Therefore, future management research should target hydro-peaking effects on terns and survival rates of both avian species to decrease the standard errors associated with each input parameter.

As resources managers are usually limited with time and funding, prioritizing research needs through sensitivity analyses may yield more efficient allocation of funds to research projects.

My work resulted in a population model with a short run time of several minutes that projects responses of terns and plovers to differences between two habitat areas on the LPR. While the model assumes avian species have the same nest and fledgling success on sandbars as they do on sandpits, if further research and monitoring detects differences in the input parameters, including discharge and various anthropogenic

affects, between the two habitat types this information can be readily incorporated. 77

However, as my assumption about the relative productivity in the two habitat types is supported by previous research (Kirsch, 1996), the model remains a valid tool for projecting the consequences of management alternatives (Deines et al., 2007). These quantitative projections can be used in a structured method to prioritize and select the

78 best alternative(s) from a set of multiple management alternatives for implementation

(Possingham, 2000).

MODEL APPLICATIONS

The population model is ideal for an adaptive management framework where the model will be constantly updated, revised, and improved through continued monitoring.

This will result in increased model accuracy and an adaptive model that responds to a changing ecosystem. However, there are no direct federal mandates for the LPR regarding terns and plovers, such as recovery objectives that are federally defined for the

Missouri River and the CPR populations (USFWS, 1990). Consequently, this has resulted in multiple agencies managing the river independent of other management agencies and no formal management plan covering the entire stretch of the LPR. As discussed and defined by Williams et al. (2007), adaptive management consists of two primary conditions: a well-defined problem and institutional commitment to the process.

While there appears to be a well-defined problem, which is the basis of my population model, there is not a central institution committed as the decision authority for the entire

LPR. However, the absence of a central decision authority should not cause multiple

resources management agencies on the LPR to abandon the adaptive management 78

process. Instead resources management agencies could collaborate, pooling institutional resources and utilizing the population model for its intended use: to pursue a conclusion to the proposed problem of resource allocation for the recovery of Least Tern and Piping

Plover populations on the LPR.

NEXT STEPS

Continued effort contributed towards furthering model advancements includes continuously updating the model with monitoring data as it becomes available. Given the small data set available on terns and plovers of the LPR that began in 2008, as future monitoring data becomes available and more input parameters transfer to being specific to the LPR region, the model will increase in accuracy and become more specialized for the LPR. For further refinement in the model, I suggest dividing the on-channel habitat into two reaches or segments. This is due to the potential differences between the upper and lower reaches of the LPR where river morphology is affected by incoming water from the Loup River in the upper reach and Elkhorn River in the lower reach. As hydro- peaking effects on habitat in the LPR for Least Tern and Piping Plover are disputed, the upper reach may contain less suspended sediment and may be more strongly affected by hydro-peaking due to the activities of the Loup River Power District on the Loup River

(Graf, 2001; Parham, 2007). However, as there are minimal mechanical alterations of the

Elkhorn River, such as dams, bank stabilization, and channel deepening for navigation, water entering the LPR from the Elkhorn may enter with a more natural hydrology and

contain more sediment than that of the Loup River. Potential differences in 79

sedimentation within the water column and hydro-peaking effects may lead to differences in sandbar development and erosion, thus altering the area of available habitat and finally affecting the avian populations.

Currently, all plausible reasons for their differing responses are summarized under productivity and density estimates. Future research should focus on the differences

80 between the two avian species responses to habitat changes. Additionally, density- dependence complicates management for both terns and plovers. During high population densities, the productivity decreases while at low densities, the productivity increases.

The density-dependent responses developed in the population model are supported by the work of Kirsch (1996) studying terns on the LPR, Catlin (2009) for plovers on the

Missouri River, and Cohen et al. (2007) for plovers on coastal islands. Kirsch (1996) suggested greater density-dependent responses on the sandpits than on sandbars (i.e., sandbars may support higher densities of birds than sandpits). During years of high flows that result in continuous sandbar flooding, more birds may nest on sandpits. However, given the higher degree of density dependence on sandpits, larger numbers of birds nesting on sandpits would lead to a greater decrease in fledglings per pair. If management disregards density-dependence, habitat construction programs may lead to ecological traps where attractions of large numbers of birds results in increased density with reductions in productivity (Gates and Gysel, 1978).

Sandbar area available for tern and plover breeding and nesting is dependent on rates of erosion, sedimentation, and vegetation, but sandpit area is more closely

dependent on various human activities, such as development and mining operations. 80

Incorporating more aspects of the two habitats would allow for increased accuracy in the estimation of projected available habitat. Additionally, monitoring the nesting success and fledgling survival at each habitat type and incorporating habitat specific rates into the model will allow further differentiation between to the two habitats. Integrating more population dynamics that may differ between the two habitats, leads to improved

81 assessment of the various management consequences and predictions of birds’ responses to the alteration and restoration of the two habitats.

MODEL CONCLUSIONS

I advise management agencies to develop an adaptive management plan that incorporates iterative decision-making, monitoring, and the population model to recover the endangered Least Tern and threatened Piping Plover on the LPR. Although there are multiple definitions of adaptive management, a recurrent theme in all adaptive management approaches is the ongoing monitoring of measurable objectives while also implementing selected actions (Walters and Holling, 1990; Field et al., 2004; Gerber et al., 2005; McFadden et al., 2011). In developing an adaptive management plan, it is important for managers to notice that, given the results from the population model, the two avian species may survive well as long as they have sufficient access to one type of habitat. However, when both habitats are lost and model projections expand further than

10 years, the populations plummet. Additionally, differences in productivity estimates and survival rates between the two species contribute largely to their specific responses to habitat changes. Therefore, it is vital to examine potential effects of management actions

further into the future for an improved understanding of population responses. 81

By utilizing this complete model description, managers gain a thorough understanding of model components and interactions between input and output parameters. The population model is a useful tool to aid resources managers in measuring the annual status of the two avian species and is best suited for a continuous process of reducing uncertainty through adaptive management. With active learning and

82 continuous monitoring, uncertainty decreases and forecast management outcomes are more easily predicted (Walters, 1986, 1997). Therefore, through a long-term monitoring program supported by the long-term population model and guided by clear objectives, the uncertainty placed on model parameters will continue to decrease while the model projections improve. This allows for more informed decision making as the number of iterations increase in the adaptive management process. Through collaboration of the multiple natural-resources agencies managing the LPR and pursuit of iterative decision- making strategies, development of measurable objectives is expected to result in a clear adaptive management plan that utilizes the population model for the recovery of Least

Terns and Piping Plovers.

LITERATURE CITED

Adolf, S.L., K. Higgins, C. Kruse, G. Pavelka, 2001. Distribution and productivity of

Least Terns and Piping Plovers on the Niobrara River. Proceedings of the South

Dakota Academy of Science 80, 231–245.

Bental, R., 1982. Nebraska’s Platte River, graphical analysis of flows. Conservation and

Survey Division, University of Nebraska, Lincoln, Nebraska, Nebraska Water

Survey Pap. 53, 47 p.

Brown, M.B., J.G. Jorgensen, 2009. 2009 Interior Least Tern and Piping Plover

monitoring, research, management, and outreach report for the Lower Platte

River, Nebraska. Tern & Plover Conservation Partnership and the Nebraska Game

and Parks Commission Nongame Bird Program, Lincoln, NE. available at

http://ternandplover.unl.edu (accessed November 2010).

Brown, M.B., J.G. Jorgensen, 2008. 2008 Interior Least Tern and Piping Plover

monitoring, research, management, and outreach report for the Lower Platte

River, Nebraska. Tern & Plover Conservation Partnership and the Nebraska Game

and Parks Commission Nongame Bird Program, Lincoln, NE. available at

http://ternandplover.unl.edu (accessed November 2010). 83

Carpenter, S.R., 2002. Building an ecology of the long now. Ecology 83: 2069–2083.

Caswell, H., M. Fujiwara, 2004. Beyond survival estimation: mark-recapture, matrix

population models, and population dynamics. Animal Biodiversity and

Conservation 27, 471–488.

Caswell, H., 2001. Matrix population models: construction, analysis and interpretation.

Second edition. Sinauer Associates, Sunderland, Massachusetts, USA.

Catlin, D.H., 2009. Population dynamics of Piping Plovers (Charadrius melodus) on the

Missouri River. Dissertation. Virginia Polytechnic Institute and State University,

Blacksburg, VA.

Cohen, J.B., L.M. Houghton, J.D. Fraser, 2007. Nesting density and reproductive success

of Piping Plovers in response to storm- and human-created habitat changes.

Wildlife Monographs 173, 1–24.

Colyvan, M., S. Linquist, W. Grey, P.E. Griffiths, J. Odenbaugh, H.P. Possingham, 2009.

Philosophical issues in ecology: recent trends and future directions. Ecology and

Society 14: 22 [online] URL:http://www.ecologyandsociety.org/vol14/iss2/art22/.

Deines, A., E. Peterson, D. Boeckner, J. Boyle, A. Keighley, J. Kogut, J. Lubben, R.

Rebarber, R. Ryan, B. Tenhumberg, S. Townley, A.J. Tyre, 2007. Robust

population management under uncertainty for structured population models.

Ecological Applications 17, 2175–2183.

Elliott, C.M., B.L. Huhmann, R.B. Jacobson, 2009. Geomorphic Classification of the

Lower Platte River, Nebraska: U.S. Geological Survey Scientific Investigations 84

Report 2009–5198, 29 p.

Eschner, T.R., R.F. Hadley, K.D. Crowley. 1983. Hydrologic and Morphologic Changes

in the Channels of the Platte River Basin in Colorado, Wyoming, and Nebraska:

A Historical Perspective. U.S. Geological Survey Professional Paper 1277-A.

Washington, DC: U.S. Government Printing Office. 39 p.

Faanes, C.A., 1983. Aspects of the nesting ecology of Least Terns and Piping Plovers in

Central Nebraska. Prairie Naturalist 15, 145–154.

Field, S. A., A.J. Tyre, N. Jonzen, J.R. Rhodes, H.P. Possingham, 2004. Minimizing the

cost of environmental management decisions by optimizing statistical thresholds.

Ecology Letters 7, 669–675.

Gates, J.E., L.W. Gysel, 1978. Avian nest dispersion and fledging success in field-forest

ecotones. Ecology 59, 871–883.

Gerber, L. R., M. Beger, M.A. McCarthy, H.P. Possingham, 2005. A theory for optimal

monitoring of marine reserves. Ecology Letters 8, 829–837.

Graf, W.L., 2001. Damage Control: Restoring the physical integrity of America’s rivers’.

Annals of the Association of American Geographers 91, 1–27.

Greeuw, S.C.H., M.B.A. van Asselt, J. Grosskurth, C. A. M. H. Storms, N. Rijkens-

Klomp, D. S. Rothman, and J. Rotmans, 2000. Cloudy crystal balls. European

Environment Agency, Copenhagen, Denmark.

Gregory, R.S., R.L. Keeney, 2002. Making Smarter Environmental Management

Decisions. Journal of the American Water Resources Association 38, 1601–1612.

Hall, L.S., P.R. Krausman, M.L. Morrison, 1997. The habitat concept and a plea for 85

standard terminology. Wildlife Society Bulletin 25, 173–182.

Hodges, K.E., 2008. Defining the problem: terminology and progress in ecology.

Frontiers in Ecology and the Environment 6, 35–42.

Jenniges, J.J., R.G. Plettner, 2008. Least Tern nesting at human created habitats in central

Nebraska. Waterbirds 31, 274–282.

Joeckel, R.M., G.M. Henebry, 2008. Channel and island change in the lower Platte River,

eastern Nebraska, USA: 1855–2005. Geomorphology doi:

10.1016/j.geomorph.2008.04.016.

Kirsch, E.M., 2000. Nebraska’s Least Terns and Piping Plovers 1987-1999: Final report.

USGS, Upper Midwest Environmental Sciences Center, 24 p.

Kirsch, E.M., 1996. Habitat selections and productivity of Least Terns on the Lower

Platte River, Nebraska. Wildlife Monographs 132, 3–48.

Larson, M.A., M.R. Ryan, B.G. Root, 2000. Piping Plover survival in the Great Plains: an

updated analysis. The Journal of Field Ornithology 71, 721–729.

Licht, D.S., 2001. Relationship of hydrological conditions and populations of breeding

piping plovers. The Prairie Naturalist 33, 208–219.

Marcus, J.F., J.J. Dinan, R.J. Johnson, E.E. Blankenship, J.L. Lackey, 2007. Directing

nest site selection of Least Terns and Piping Plovers. Waterbirds 30, 251–258.

Massey, B.W., D.W. Bradley, J.L. Atwood, 1992. Demography of a California Least

Tern colony including effects of the 1982-1983 El Nino. The Condor 94, 976–

983.

McFadden, J.E., T.L. Hiller, A.J. Tyre, 2011. Evaluating the Efficacy of Adaptive 86

Management Approaches: Is there a Formula for Success? Journal of

Environmental Management 92, 1354–1359.

Melbourne, B.A., A. Hastings, 2008. Extinction risk depends strongly on factors

contributing to stochasticity. Nature 454, 100–103.

Microsoft, 2007. Microsoft Excel [computer software]. Redmond, Washington:

Microsoft.

NRC (National Research Council). 2005. Endangered and threatened species of the Platte

River, National Academies Press, Washington, D.C.

Nicolson, C.R., A.M. Starfield, G.P. Kofinas, J.A. Kruse, 2002. Ten heuristics for

interdisciplinary modeling projects. Ecosystems 5, 376–384.

Owen-Smith, N., 2007. Introduction to modeling in wildlife and resource conservation.

Blackwell Publishing, Malden, MA, USA.

Palmer, M.A., 2009. Reforming watershed restoration: science in need of application and

applications in need of science. Estuaries and Coasts 32, 1–17.

Parham, J.E., 2007. Hydrologic analysis of the lower Platte River from 1954-2004, with

special emphasis on habitats of the Endangered Least Tern, Piping Plover, and

Pallid Sturgeon. Prepared for Nebraska Game and Parks Commission, Lincoln,

NE.

Parma, A.M., NCEAS working group on population management, 1998. What can

adaptive management do for our fish, forests, food, and biodiversity? Integrative

Biology 1, 16–26. 87

Possingham, H.P., S.J. Andelman, B.R. Noon, S. Trombulak, H.R. Pulliam, 2001.

Making smart conservation decisions. In: Soule ME, Orians GH (eds) Research

priorities for nature conservation. Island Press, Washington, DC, p 225–244.

Possingham, H.P., 2000. Is Environmental Research a Waste of Time? Ecological

Management & Restoration 1, 81–82.

R Development Core Team (2009). R: A language and environment for statistical

computing. R Foundation for Statistical Computing, Vienna, Austria. Available

from http://www.R-project.org (accessed November 2010).

Renken, R.B., J.W. Smith, 1995. Annual adult survival of Interior Least Terns. Journal of

Field Ornithology 66, 112–116.

Rodekor, D.A., K.W. Engelbrecht, 1988. Island and bank morphological changes

detected in the Platte River bounding the Papio Natural Resources District from

1949 through 1988. Center for Advanced Land Management and Information

Technologies, University of Nebraska–Lincoln, Nebraska.

Rubin, E.S., W.M. Boyce, E.P. Caswell-Chen, 2002. Modeling demographic processes in

an endangered population of Bighorn Sheep. Journal of Wildlife Management 66,

796–810.

Schweitzer, S.H., D.M Leslie, JR., 1999. Nesting habitat of Least Terns (Sterna

antillarum athalassos) on an Inland Alkaline Flat. American Midland Naturalist

142, 173–180.

Thompson, B.C., J.A. Jackson, J. Burger, L.A. Hill, E.M. Kirsch, L.J. Atwood, 1997.

Least tern (Sterna antillarum). The birds of North American, number 290. The 88

American Ornithologists’ Union, Washington, D.C., USA, and The Academy of

Natural Science, Philadelphia, Pennsylvania, USA.

U.S. Fish and Wildlife Service. 1985a. Endangered and threatened wildlife and plants:

Interior population of the least tern determination to be endangered. Federal

U.S. Fish and Wildlife Service. 1985b. Determination of endangered and threatened

status for the piping plover. Federal Register 50, 50720-50734.

USFWS (U.S. Fish and Wildlife Service). 1990. Interior population of the least tern,

Sterna antillarum, recovery plan. U.S. Fish and Wildlife Service, Twin Cities,

Minnesota, USA.

Walters, C., 1997. Challenges in adaptive management of riparian and coastal

ecosystems. Conservation Ecology 1, 1. http://www.consecol.org/vol1/iss2/art1.

Accessed 30 March 2010.

Walters, C. J., C.S. Holling, 1990. Large-scale management experiments and learning by

doing. Ecology 71, 2060–2068.

Walters, C.J., 1986. Adaptive management of renewable resources. MacMillan, New

York, USA.

Williams, B. K., R.C. Szaro, C.D. Shapiro, 2007. Adaptive Management: The U.S.

Department of the Interior Technical Guide. Adaptive Management Working

Group, U.S. Department of the Interior, Washington, D.C.

Ziewitz, J.W., J.G. Sidle, J.J. Dinan, 1992. Habitat Conservation for Nesting Least Terns

and Piping Plovers on the Platte River, Nebraska. Prairie Naturalist 24, 1–20. 89

Chapter 4: ASSESSING COMPETING RESEARCH HYPOTHESES FOR

EXPERIMENTATION THROUGH ADAPTIVE MANAGEMENT ON THE

CENTRAL PLATTE RIVER, NEBRASKA

Abstract:

The central Platte River (CPR), Nebraska, is a complicated ecosystem where natural-resources management decisions affect endangered and threatened species such as the Interior Least Tern (Sternula antillarum athalassos) and Piping Plover (Charadius melodus). Portions of the CPR are managed under the Platte River Recovery

Implementation Program (hereafter simply Program) for four federally endangered and threatened species. The Program is pursuing management of the endangered and threatened species of the river through an adaptive management based process. Because high uncertainty is associated with the responses of these species to habitat restoration and other resource uses, I tested alternative hypotheses using a multi-model analysis framework based on simulated data to simplify hypotheses and prioritize research and management needs. I developed a model set with 10 models and applied it to five

simulated scenarios using three analysis methods that (overdispersion) differed by family 90

distribution and the type of Akaike’s Information Criterion. As the analysis focused on the Interior Least Tern responses to ecological factors, model results suggest that in not accounting for overdispersion in the data leads to a greater probability of concluding that something is falsely affecting the response variable when the parameter effect sizes are close to 0. In applying the Shannon-Weiner Index of diversity to model weights of all five simulated data scenarios, the results suggest that as model weight diversity increases,

91 multi-model inference strength decreases. Thus, by utilizing statistical models for evaluating management consequences, iterative decision-making will allow for continuous updating of models, as more monitoring data becomes available, influencing future management decisions. The process of evaluating effects of ecological factors is helpful in setting and prioritizing objectives and implementing actions for adaptively managing complicated ecosystems.

INTRODUCTION

The central Platte River (CPR) in Nebraska, once with wide and shallow braided channel morphology, now has stabilized banks, deeper channels, and increased flow

(Zallen, 1997; Kenney, 2000; Echeverria, 2001). Alterations to riverine habitat over the last several decades resulted from water diversions, land-use changes, and other basin alterations and contributed to the listing of various species (Sinokrot and Gulliver, 2000;

Kroeger and McMurray, 2008; Smith, 2011). Portions of the CPR, one of the largest

Great Plains riverine ecosystems, are managed by Headwaters Corporation through their

Platte River Recovery Implementation Program (hereafter simply Program) for four federally endangered and threatened species (Schneider et al., 2005). The Program pursues management of endangered and threatened species of the river through an adaptive management based process (Platte River Recovery Implementation Program,

2008).

Adaptive management provides an improved method for incorporation of research into management decisions where uncertainty places limitations on contributions of science (Reynolds et al., 1996; Lee and Bradshaw, 1998; Berg et al., 1999; Robertson

and Hull, 2001). When adaptively managing complicated systems, utilizing structured 92

decision-making provides managers with a formal process for considering all aspects of the complicated decisions they face (Gregory and Keeney, 2002). By repeating decisions within a structured decision-making approach, natural-resources managers learn through an ongoing process of implementing various management actions, monitoring management outcomes, and updating ecological models by comparing actual outcomes

93 with expected outcomes (Hilborn and Walters, 1981; Walters, 1986; Williams, 1996;

Carpenter, 2002; Johnson et al., 2002). In addition to emphasizing experimentation as a method of rapidly improving knowledge of the system (Walters, 1986; Williams et al.,

2007), the Program pursues decision-making to link management and science through engaging experts on species recovery, river management, and policy. Management objectives focus on improving production of the Interior Least Tern (Sterna antillarum athalassos) and Piping Plover (Charadrius melodus), improving survival of Whooping

Cranes (Grus americana) during migration, avoid adverse impacts on Pallid Sturgeon

(Scaphirhynchus albus), and reduce likelihood of future species listings (Marmorek et al.,

2009).

Although much research has been conducted on the CPR, there remains a high level of scientific disagreement on the habitat needs of the focus species (Supalla et al.,

2002; National Research Council, 2005). Therefore, to gain information about species’ responses to various habitat modifications for improved management towards achieving the objectives, the Program developed over forty competing research hypotheses (Platte

River Recovery Implementation Program, 2008). Such an extraordinary number of

hypotheses would require endless experimentation on the river that would be 93

unnecessarily challenging and costly for management. Multi-model analysis can reduce the number of competing hypotheses by combining similar hypotheses into a single set of models that can be evaluated simultaneously.

Most natural-resources managers are not privileged with unlimited time and funding, and therefore benefit greatly from efficient and effective methods that aid

94 decision-making. Multi-model inference is a statistical method of simultaneously comparing numerous hypotheses about ecological interactions within a complicated system by assessing alternative models based on given data and relative likelihoods

(Anderson et al., 2000; Burnham and Anderson, 2002). In natural-resources management, results from multi-model inference serve two primary functions, decrease uncertainty about the hypotheses and aid in prioritizing research and management projects. However, with minimal thought attributed to model development leading to incorrect model structures, model results are often vague with weak inference where multiple models in a set are plausible approximations of the collected data (Guthery et al., 2005; Smith et al., 2009).

Weak inference may originate from several other issues, such as too many models within the model set, the model structures within the set are incorrect, and a variety of methodological issues (Anderson and Burnham, 2002; Smith et al., 2009).

Methodological issues refer to limitations in the researcher’s ability to model hypotheses accurately and in understanding the entire suite of statistical methods appropriate for the given data. Additionally, weak inference results from not applying techniques, such as

model averaging, to gain further information on parameters within top ranked models. 94

Ignorance of model selection uncertainty is troublesome as it may lead to the assumption that the top ranked model is the truth. While model averaging increases knowledge gained from a given data set, information gaps inevitably remain with weak inference.

When weak inferences are encountered, adaptive management is the most relevant management recommendation especially when decisions must be made

95 regardless of the level of knowledge or uncertainty (Rehme et al., 2011). Through emphasizing stakeholder involvement, personal and political values, decision-making, uncertainty, and modeling in an adaptive management plan (Gregory and Keeney, 2002;

Williams et al., 2007), resulting active learning and continuous monitoring decreases uncertainty and forecast management outcomes are more easily predicted leading to stronger model inference (Walters, 1986, 1997). Improvements in model inference allows for more informed decision making as the number of iterations increase in the adaptive management process. Therefore, in an iterative decision-making context with a monitoring program guided by clear objectives, the uncertainty placed on model parameters will continue to decrease while model selection improves, thus influencing future management decisions (Walters 1986, 1997; Carpenter, 2002). Adaptive management allows managers to continue making decisions despite uncertainty and weak inferences.

A recent study connected research with management by modeling the response of reproductive success of Least Terns and Piping Plovers on the CPR to various habitat parameters (Howlin et al., 2008). However, the study suffered from an extreme lack of

data during six years of collection as often occurs when researching endangered and 95

threatened species. Using forward selection with Akaike’s Information Criterion, analysis results were limited in providing valuable information for critical management decisions, as the data were unable to support strong inference in model selection.

Although the analysis was statistically and scientifically sound and given the challenging

96 nature of conveying statistical results, the analysis was not interpreted in such a way that was easily connected to management.

In decision-making, statistical models are used in evaluating management consequences based on current data (Possingham et al., 2001; Nicolson et al., 2002;

Williams et al., 2007). I offer a method of simplifying hypotheses and prioritizing research and management needs by providing simulated data scenarios analyzed in a multi-model analysis framework. The simulated data scenarios aid in exploring potential results of the reality on the CPR for the Interior Least Tern. The simulation may connect research and monitoring data collected by the Program on Least Terns breeding and nesting on the CPR, Nebraska. Through assuming future monitoring data, I used simulated data to compare Least Tern population numbers to various research hypotheses on the CPR. To investigate the relationship between population sizes and various ecological factors, such as bare sand area and active channel width, I consolidated research hypotheses developed by the Program into sets of statistical models for simulated multi-model analysis using generalized linear models. My final analysis objective included developing future scenarios to generate plausible monitoring data

using the number of Least Tern adults as an example. This initial process of evaluating 96

the effect sizes of various ecological factors is not a traditional power analysis, but is helpful in setting and prioritizing objectives and implementing actions for adaptively managing large river systems. Cumulatively, the simulated analysis may provide insight on the detectability of an effect and the direct implications for sampling effort.

METHODS

I used multi-model inference with generalized linear models in Program R to investigate the relationship between population size of the Interior Least Tern and various ecological factors, such as bare sand area and active channel width (R Development Core

Team, 2009). I proposed a set of 10 a priori models that incorporated six covariates derived from the available hypotheses developed by the Program (Table 1). My response variable was the number of tern adults and my six covariates included area of bare river sand, area of bare sandpit, bare sand elevation, percent vegetation cover, active channel width, and number of prey fish. Within the set of models, I included a null model (no effects) and a global model (all covariates have an effect).

Table 1. Models derived from research hypotheses developed by the Platte River Recovery Implementation Program used in multi-model analysis with Least Tern adult numbers as the response variable. Model Number Model Title Model Structure 1 (Null) Null Model Adult tern numbers ~1 2 River Sand Adult tern numbers ~ bare river sand area 3 Sandpit Adult tern numbers ~ bare sandpit area 4 Elevation Adult tern numbers ~ bare sand elevation 5 Vegetation Adult tern numbers ~ percent vegetation cover 97

8 Channel Adult tern numbers ~ active channel width 97

9 Fish Adult tern numbers ~ number of prey fish 6 Additive Adult tern numbers ~ bare river sand area + bare sand River Sand elevation + percent vegetation cover 7 Sand Area Adult tern numbers ~ bare river sand area * bare sandpit area 10 (Global) Global Adult tern numbers ~ bare river sand area + bare Model sandpit area + bare sand elevation + percent vegetation cover + active channel width + number of prey fish

I developed five simulated data scenarios that span a wide range of parameter effect sizes and include a No Effect scenario, a Tapering Baseline scenario, Similar

Effects scenario, Strong Effects scenario, and a Large Difference effects scenario (Table

2). The data scenarios provide several alternative relationships between the various habitat factors and Tern population. The effect sizes for each scenario are theoretical for comparison purposes of different elements of the research hypotheses development by the Program. The parameters within the No Effect scenario are zero and simulate a Tern population that is not affected by the selected ecological factors. I expected the best-fit

AIC model for the No Effect scenario to be the Null Model. The Tapering Baseline scenario contains parameters that diminish with each factor included. The Similar

Effects scenario contains a narrow range of diminishing parameter effect sizes where as the Strong Effects scenario contains larger parameter effect sizes and the Large

Difference scenario contains a wide range of strong, yet diminishing parameter effect sizes. The Strong Effects and Large Difference scenarios contain the strongest parameter effect sizes of all five scenarios. I expected the best-fit AIC model for the Tapering

Baseline, Similar Effects, Strong Effects, and Large Difference scenarios to be the Global

Model because each parameter has an effect regardless of size. The simulated ecological 98

factors (i.e., parameters) are based on a random number generator with identical means

(μ) and standard deviations (σ).

Table 2. Parameter effect sizes for each simulation scenario. Assigned values are not based on real data. Distributions are identical for each parameter (i.e. identical means and standard deviations).

Scenario

No Tapering Similar Strong Large Parameter Effect Effect Baseline Effect Effect Difference Bare River Sand Area 0 0.1 0.1 0.5 0.5 Bare Sandpit Area 0 -0.05 -0.08 -0.25 -0.16 Bare Sand Elevation 0 0.025 0.06 0.12 0.05 Percent Vegetation 0 -0.012 -0.047 -0.05 -0.018 Cover Active Channel Width 0 0.005 0.035 0.03 0.006 Number of Prey Fish 0 0.003 0.025 0.01 0.002

Provided the high uncertainty of tern responses to basic habitat elements of bare river sand area and bare sandpit area and emphasis placed on these two variables by the

Program, I simulated tern responses to changes in bare river sand area and bare sandpit area simultaneously (Figure 1). The underlying assumption for all scenarios, is bare sandpit area has a negative effect on terns nesting on bare river sand area. The data simulations support that adult tern numbers increase with increasing bare river sand area, but decrease with increasing bare sandpit area. Therefore, bare sandpit area has a

negative effect on the number of tern adults that are nesting on bare river sand. 99

I designated a normal distribution to all covariates with a μ of zero and a standard deviation of σ with a number of observations (n) was 100. The response variable per scenario ( ), number of tern adults ( ), had a negative binomial distribution ( ) where n = 100 with a mean (μT ) as a function of the covariates with effect sizes specific to the simulated scenarios using the following equations:

100 where is the parameter effect size for the area of bare river sand specific to scenario

( ), is the bare river sand area covariate, is the parameter effect size for the area of bare sandpit, is the bare sandpit area covariate, is the bare sand elevation parameter effect size, is the bare sand elevation covariate, is the parameter effect size for percent vegetation cover, is the percent vegetation cover covariate,

is the channel width parameter effect size, is the channel width covariate, is the number of prey fish parameter effect size, is the number of prey fish covariate, and is the normal distribution with n = 100, μ = 0, and σ = 0.01. I used the following equation to estimate the number of tern adults per scenario:

100

Figure 1. Underlying assumption for simulated data scenarios where bare sandpit area has a negative effect on the number of tern adults nesting on bare river sand. Shown is the assumption specific to parameter effect sizes in the Strong Effect data scenario where white indicates an increase in tern population size and red indicates a decrease in tern population size.

101

For the multi-model analysis, I used Akaike’s Information Criterion (AIC) to determine initial model ranking and Akaike weights (wi) to determine the level of supportive evidence each model in each simulated data scenario (Burnham and Anderson,

2002). The model set was identical for each simulated scenario. To examine the effects of overdispersed data on the multi-model inference results, I conducted three analyses that differed by distribution and type of AIC analysis used: a Poisson distribution with

AIC (PA method), a Poisson distribution with Quasi-AIC (PQ method), and a Negative

Binomial distribution with AIC (NA method). I applied the distributions to the model sets for each simulated scenario. The PA analysis did not offer any methods for accounting for overdispersion and served as the control method. I used the Poisson distribution because the simulated data was considered count data and the Negative

Binomial distribution for the third analysis method as another solution for overdispersed data (Lindsey, 2004). Lastly, I used Quasi-AIC as it incorporates an overdispersion parameter, allowing for another method of accounting for overdispersion (Zuur et al.,

2007).

To examine the variation among model ranking, I estimated the probability of

101

101 each model when ranked as the best AIC model over 100 iterations using the average

relative AIC (ΔAIC). To clarify the frequency results, I applied Ivlev’s Index of

Electivity ( ) to the frequency estimates for analysis method and simulation using the following equation (Ivlev, 1961):

102 where is the relative probability estimate and is the relative availability of model ranking in the set of models. With a scale of 1.0 to -1.0, if and are equal for all models (i.e., 0.1) where , model ranking is random. If and are different for a model, depending on the direction of deviation from 0, model ranking for a particular model is selected for or against.

In assessing model-inference strength, I applied the Shannon-Wiener Diversity

Index to model wi of all five simulated data scenarios. I used the following equation to obtain the proportional model wi ( ) using the following equation:

where is the wi for a given model . In applying the diversity index, I used the following equation:

where is the model weight diversity. A maximum possible diversity occurs when all models are weighted identically. I expected that when is high, model-inference strength is low.

102

102 I estimated a recommended sample size for all five data simulation scenarios by

running the multi-model analysis for the each scenario with different sample sizes for 100 iterations each. Tested sample sizes included n = 50, n = 100, n = 200, and n = 500. To detect a final relationship between sample size and , I treated for each test as a function of sample size. I expect that increases in sample size will not result in a decrease in Hs for the No Effect scenario.

103

RESULTS

To understand model output resulting from theoretical effect sizes that I used to compare the different elements of the hypotheses, I compared adult tern numbers to the variation in the effect size of bare river sand area between scenarios. The parameter effect size of bare river sand area was relatively low (i.e. 0 or 0.1) for the No Effect,

Tapering Baseline, and Similar Effects scenarios and resulted in minimal to no adult tern response for one example from one of the 100 iterations (Figures 2A-C). However, as the parameter effect size of bare river sand area was larger (i.e. 0.5) for the Strong Effects and Large Difference scenarios, the simulated data suggests a positive trend between adult tern numbers and area of bare river sand as expected (Figures 2D & E).

103

104

Figure 2. Simulated results of the tern population size in response to bare river sand area for a single iteration of each simulation scenarios, including the No Effect (A), Tapering Baseline (B), Similar Effects (C), Strong Effects (D), and Large Difference (E).

AIC MODEL SELECTION

I simulated the AIC analysis for 100 iterations for each scenario and analysis method. As an example of one iteration for the Strong Effects scenario, I found

105 differences among model ranking between analysis methods (Table 3). Model ranking for the PA method suggests 86% of the model weight supports bare sandpit area as the best predictor of adult tern numbers followed by four models with single parameters. In contrast, when utilizing methods that account for overdispersion, the PQ and NA methods suggest similar weights of 52% support the interaction of bare river sand area and bare sandpit area as the best predictor. Although the null model was ranked higher in the PA method than for the PQ and NA methods, the global model was ranked as the second or third best model in the PQ and NA methods. It appears that analysis methods that account for overdispersion support examination of more complex models. However, the total weight of the best model decreased in the overdispersion methods. Refer to

Appendix A for average AIC results over 100 iterations for all scenarios and analysis methods.

Table 3. Results for generalized linear model selection for a single iteration the Strong Effect scenario computed with a Poisson distribution and Akaike’s Information Criterion, a Poisson distribution and Quasi-AIC, and a Negative Binomial distribution and AIC using Program R for model evaluation.

a b c d e f Model Structure AIC /QAIC K ΔAIC /ΔQAIC wi 105

105 PAg

3* Psandh 442.54 7 0 0.86 8 Chwidi 446.21 4 3.67 0.14 1 Null model 453.96 4 11.42 0 2 Rsandj 454.35 2 11.81 0 6 Rsand + elevak + 469.58 2 27.04 0 vegcol 9 Pfishm 477.23 2 34.68 0 7 Rsand * psand 480.05 1 37.51 0 5 Vegco 480.89 2 38.35 0 4 Eleva 481.18 2 38.63 0

106

Table 3. continued.

Model Structure AIC/QAIC K ΔAIC/ΔQAIC wi 10 Rsand + psand + 481.89 2 39.35 0 eleva + vegco + chwid + pfish

PQn 7 Rsand * psand 229.81 4 0 0.52 2 Rsand 231.49 2 1.68 0.22 10 Rsand + psand + 231.92 7 2.11 0.18 eleva + vegco + chwid + pfish 6 Rsand + eleva + 233.68 4 3.88 0.07 vegco 3 Psand 239.11 2 9.3 0 5 Vegco 242.93 2 13.12 0 1 Null model 243.22 1 13.41 0 9 Pfish 244.76 2 14.96 0 8 Chwid 244.91 2 15.1 0 4 Eleva 245.26 2 15.46 0

NAo 7 Rsand * psand 423.29 5 0 0.52 10 Rsand + psand + 424.43 8 1.15 0.29 eleva + vegco + chwid + pfish 2 Rsand 426 3 2.71 0.13

6 Rsand + eleva + 427.94 5 4.65 0.05

vegco 106

3 Psand 431.48 3 8.19 0.01 106

5 Vegco 435.78 3 12.5 0 1 Null model 435.89 2 12.6 0 9 Pfish 437.37 3 14.09 0 8 Chwid 437.52 3 14.24 0 4 Eleva 437.81 3 14.52 0 a-f AIC = Akaike’s Information Criterion; QAIC = Quasi-AIC; K = number of model parameters; ΔAIC = relative AIC; ΔQAIC=relative Quasi-AIC for overdispersion; wi = Akaike weight g, n, o PA = Poisson Distribution and AIC analysis method; PQ = Poisson Distribution and QAIC analysis method; NA = Negative Binomial Distribution and AIC analysis method. h-mPsand = bare sandpit area; chwid = active channel width; rsand = bare river sand area; eleva = bare sand elevation; vegco = percent vegetation cover; pfish = number of prey fish *Number of each model as they appear in the model set

107

MODEL RANKING

The null model was ranked as the best model most often as suggested by Ivlev’s

Index of Electivity for the No Effects scenario over 100 iterations under the PQ and NA methods (Figure 3A). All other models received similar ranking, except for the global model that was selected against in the PQ and NA methods with an = -1.0. With the

PA method, all models, including the null and global models, received similar model rankings with an near zero. Therefore, the PQ and NA methods lend most support to the “true” model, which is the null model for the No Effects scenario, than the PA method.

Results were similar for the Tapering Baseline and Similar Effects scenarios

(Figure 3B & C). While all parameters had an effect on the number of adult terns in the two scenarios, the global model was selected against with a strong negative for the PQ and NA methods (Tapering Baseline: = -1.0 for PQ, = -0.82 for NA; Similar

Effects: = -0.82 for PQ, = -0.54 for NA). However, with the PA method, the global model was ranked as the top model more often under this method for the Similar Effects

scenario ( = 0.35). Therefore, the PA method lends most support to the global model,

107

107 the expected model, for the Tapering Baseline and Similar Effects scenario than the PQ

and NA methods.

Results were similar for the Strong Effects and Large Difference scenarios

(Figure 3D & E). As all parameters had a stronger effect on the number of adult terns in the two scenarios than in the Tapering Baseline and Similar Effects scenarios, the global model was ranked more highly in the Strong Effects and Large Difference scenarios. The

108

PA method ranked the global model highest of the three methods with an = 0.75 for the Strong Effects scenario and = 0.63 for the Large Difference scenario. The methods for overdispersion rank the global model higher in the Strong Effects scenario than in the Large Difference scenario, the for the PQ and NA methods is lower than the PA method for both scenarios. However, for the Strong Effects and Large Difference scenarios, model-inference appears to be stronger for the two scenarios in comparison to the Tapering Baseline and Similar Effects scenarios. Refer to Appendix A for more detailed of model ranking estimates.

Results for the model weight diversity (Hs) estimates suggest that the No Effects scenario had the highest Hs of all scenarios with a mean of 1.71 (Figure 4). The Tapering

Baseline and Similar Effects scenarios had similar Hs with a mean of 1.46 and 1.38 respectively. The Large Difference had a lower Hs with a mean of 0.62. The Strong

Effects scenario had the lowest Hs with a mean of 0.31.

RECOMMENDED SAMPLE SIZE

The power analysis used to determine the recommended sample size suggests that

as sample size increases, Hs decreases for all scenarios except the No Effect (Figure 5).

108

108 For the No Effect scenario, regardless of the number of samples collected, model

inference strength does not change (Figure 5A). Results for the Tapering Baseline and

Similar Effects scenarios suggest that model inference strength is unlikely to increase until N = 500 (Figures 5B & 5C). When effect sizes increase, as in the Strong Effects and Large Difference scenarios, model inference strength appears to increase with increasing sample size (Figures 5D & 5E).

109

Figure 3. Estimated frequencies for each model ranked as the best Akaike’s Information Criterion (AIC) model out of 100 iterations using Ivlev’s Index for Electivity. Shown are results for the No Effect (A), Tapering Baseline (B), Similar Effects (C), Strong Effects (D), and Large Difference (E) data scenarios using the Poisson distribution with AIC (orange cirlces), Poisson distribution with Quasi-AIC (green squares), and the Negative Binomial distribution with AIC (blue triangles) analysis methods. The model sets were consistent for each scenario and analysis method.

110

Figure 4. Cumulative probability results for the model weight diversity (Hs) for each simulated data scenario over 100 iterations using the Poisson distribution and AIC analysis method as an example. Shown is the Hs for the No Effect Scenario (red line), Tapering Baseline scenario (light blue line), Similar Effects scenario (green line), Strong Effects scenario (black line), and Large Difference scenario (light blue).

110

111

Figure 5. Results for power analysis of the No Effect (A), Tapering Baseline (B), Similar Effects (C), Strong Effects (D), and Large Difference (E) scenarios showing model weight diversity (Hs) as a function of sample size (N). Shown is the mean (horizontal line in center for each box), 25th and 75th percentiles (ends of boxes), 10th and 90th percentiles (vertical lines), and outliers (open circles) over 100 iterations for each sample size test.

112

DISCUSSION

Overall, the simulation analysis suggests that as the parameter effect sizes increase, model ranking switches from supporting simpler models to higher ranking of more complex models for all three analysis methods. As this analysis was simulated with a large sample size (n = 100), multi-model inference was weak for the scenarios with small effect sizes including the Tapering Baseline and Similar Effects scenarios.

Therefore, when effect sizes vary, model ranking differs where the required sample size needed to detect strong effects is smaller compared to sample sizes needed to detect weak effects. Additionally, model ranking differed between analysis methods. The results suggested that when effect sizes are small or 0, PQ and NA methods selected the “true” model more often than the PA method. Between analysis methods and data scenarios, it appears that as effect sizes increase, the impacts of overdispersion affecting the selection of the “true” model, decrease. Therefore, not accounting for overdispersion may lead to a greater probability of falsely concluding that a relationship exists when parameter effect sizes are close to zero.

The simulated data scenarios aid in exploring potential results of the reality on the

112

CPR for the Interior Least Tern in response to several environmental factors. The data

scenarios improve the understanding of monitoring data a priori to implementing a management action. Through assuming various scenarios of potential monitoring data collected over the next ten years, natural-resources managers gain information about an uncertain system and decrease the likelihood of future surprises. Additionally, simulations may aid in quantifying the value of learning for a given adaptive

113 management project (Probert et al., 2010). When applying the power analysis as described above, obtaining estimates for recommended sample sizes aid managers in determining which sample size meets the needs of their objectives best. With this method, there is not one correct sample size suggested, but rather the best sample size is dependent on the value placed on learning and various constraints, such as time and funding. By exploring various monitoring techniques and values of learning as an objective, estimates of knowledge-based objectives provide natural-resources managers with a method of evaluating the utility of experimentation.

Natural-resources managers are often confronted with high uncertainty and funding limitations where timely decisions are essential. Multi-model inference aids in coping with the dilemmas of uncertainty and time by decreasing uncertainty about testable hypotheses and aiding in prioritizing research and management projects. A priori dedication to model development is imperative to achieve strong inference and maximize learning. However, with poorly developed model analyses, small sample sizes, and methodological issues statistical analyses often result in weak inference (Anderson and

Burnham, 2002; Smith et al., 2009; Rehme et al., 2011). Therefore, when natural-

113

113 resources managers encounter problems requiring timely decisions with high uncertainty

resulting from weak inference caused by a multitude of plausible sources, adaptive management is the most relevant management recommendation. Adaptive management provides clear connections between available science and stakeholder values by developing values into measurable objectives that drive iterative research and monitoring

114 plans. Through this iterative process, managers continue making structured decisions driven by values and measured by data that decrease uncertainty with each iteration.

Without measurable objectives, science is limited in effectively influencing management. Likewise, management is limited to personal and political values without access to science. As the connection between ecologists and managers is vital for decreasing uncertainty, I provide analyses of five data simulation scenarios focused on various realities of breeding and nesting Least Terns on the CPR, Nebraska. The scenarios provide a scientific method of evaluating potential realities in connecting various hypotheses to management efforts, thus decreasing surprises of future events as management is prepared for multiple potential outcomes.

For effective adaptive management, research hypotheses must be testable and falsifiable. However, management complications arise for many wildlife population managers due to environmental variation and agency time constraints (Theberge et al.,

2006). Due to these two complications, obtaining falsifiable test results becomes more challenging, especially if the hypotheses or objectives are not measurable. Measurable hypotheses and objectives focus on population vital rates, such as survival probabilities.

114

I advise a priori development of analysis methods for the prioritization of research

hypotheses as developed by the Program. With a large number of hypotheses and inescapable constraints (i.e., time, ecological stochasticity, managing endangered species), higher prioritization should be placed on research projects focused on testable and falsifiable hypotheses (i.e., survival rates and strong effect sizes). Additionally, prioritizing research hypotheses that focus on stronger parameter effect sizes will yield

115 stronger model inference than weak effect sizes given limited monitoring efforts. With thorough evaluation of research hypotheses and expected parameter effect sizes, the

Program is likely to maximize learning opportunities while proceeding with iterative decision-making. In maximizing learning, the Program will continue to decrease uncertainty about responses of Least Terns to various ecological factors on the CPR,

Nebraska, through adaptive management.

115

116

LITERATURE CITED

Anderson, D.R., K.P. Burnham, 2002. Avoiding pitfalls when using information-theoretic

methods. The Journal of Wildlife Management 66, 912–918.

Anderson, D.R., K.P. Burnham, W.L. Thompson, 2000. Null hypothesis testing:

Problems, prevalence, and an alternative. Journal of Wildlife Management 64,

912–923.

Berg, J., B. Bradshaw, J. Carbone, C. Chojnacky, S. Conroy, D. Cleaves, R. Solomon, S.

Yonts-Shepherd, 1999. Decision Protocol 2.0. U.S. Forest Service FS-634.

Burnham, K.P., D.R. Anderson, 2002. Model selection and multi-model inference: a

practical information-theoretic approach. Springer-Verlag, New York.

Carpenter, S.R., 2002. Ecological futures: building an ecology of the long now. Ecology

83, 2069–2083.

Echeverria, J.D., 2001. No success like failure: the Platte River collaborative watershed

planning process. William and Mary Environmental Law and Policy Review 25,

559–603.

Gregory, R.S., R.L. Keeney, 2002. Making Smarter Environmental Management

116

Decisions. Journal of the American Water Resources Association 38, 1601–1612.

Guthery, F.S., L.A. Brennan, M.J. Peterson, J.J. Lusk, 2005. Information theory in

wildlife science: Critique and viewpoint. Journal of Wildlife Management 69,

457–465.

Hilborn, R., C.J. Walters, 1981. Pitfalls of Environmental Baseline and Process Studies.

Environmental Impact Assessment Review 2, 265–278.

117

Howlin, S., D. Strickland, C. Derby, 2008. Interior Least Tern and Piping Plover

Reproductive Monitoring During the Cooperative Agreement (2001-2006),

Central Platte River, Nebraska. Prepared for the Platte River Recovery

Implementation Program. 16 p.

Ivlev, V.S., 1961. Experimental ecology of the feeding of fishes. Yale University Press,

New Haven, Connecticut.

Johnson, F A., W.L. Kendall, J.A. Dubovsky, 2002. Conditions and limitations on

learning in the adaptive management of mallard harvests. Wildlife Society

Bulletin 30, 176–185.

Kenney, D.S., 2000. Arguing about consensus: Examining the case against western

watershed initiatives and other collaborative groups active in natural resources

management. Boulder, CO: Natural Resources Law Center, University of

Colorado School of Law.

Kroeger, T., A. McMurray, 2008. Economic benefits of conserving natural lands: case

study: Central Platte biologically unique landscape, Nebraska. Prepared for the

Doris Duke Charitable Foundation. Washington, DC: Defenders of Wildlife. 47p.

117

Lee, D.C., G.A. Bradshaw, 1998. Making monitoring work for managers. U.S. Forest

Service. Available at: http://www.icbemp.gov/spatial/lee_monitor/begin.html

(Accessed March 2010).

Lindsey, K.Y., 2004. Introduction to applied statistics. A modeling approach, 2nd edition.

Oxford, UK: Oxford University Press.

118

Marmorek, D., P. Dixon, D. Galat, R. Jacobson, K. Loftin, J. Nestler, 2009. Independent

Scientific Advisory Committee 2009 Report to the Governance Committee on the

Platte River Recovery Implementation Program. Platte River Recovery

Implementation Program, Nebraska. Available from:

http://www.platteriverprogram.org (Accessed December 2010).

National Research Council (U.S.), 2005. Endangered and threatened species of the Platte

River. Washington, D.C., National Academies Press.

Nicolson, C.R., A.M. Starfield, G.P. Kofinas, J.A. Kruse, 2002. Ten heuristics for

interdisciplinary modeling projects. Ecosystems 5, 376–384.

Platte River Recovery Implementation Program, 2008. Prepared by Executive Director’s

Office, 2008. Strategic science plan for adaptive management plan

implementation, 2009-2013. U.S. Department of the Interior, State of Wyoming,

State of Nebraska, State of Colorado. Available from:

http://www.platteriverprogram.org (Accessed December 2010).

Possingham, H.P., S.J. Andelman, B.R. Noon, S. Trombulak, H.R. Pulliam, 2001.

Making smart conservation decisions. In: Soule ME, Orians GH (eds) Research

118

priorities for nature conservation. Island Press, Washington, DC, p 225–244.

Probert, W.J.M, C.E. Hauser, E. McDonald-Madden, M.C. Runge, P.W.J. Baxter, H.P.

Possingham, 2010. Managing and learning with multiple models: Objectives and

optimization algorithms. Biological Conservation 144, 1237–1245.

119

R Development Core Team, 2009. R: A language and environment for statistical

computing. R Foundation for Statistical Computing, Vienna, Austria. Available

from: http://www.R-project.org (Accessed November 2010).

Rehme, S.E., L.A. Powell, C. Allen, 2011. Multimodel inference and adaptive

management. Journal of Environmental Management 92, 1360–1364.

Reynolds, K., P. Cunningham, L. Bednar, M. Sanders, M. Foster, R. Olson, D. Schmolt,

D. Latham, B. Miller, J. Steffenson, 1996. A knowledge-based information

management system for watershed analysis in the Pacific Northwest U.S. AI

Applications 10, 9–22.

Robertson, D.P., R.B. Hull, 2001. Beyond biology: toward a more public ecology for

conservation. Conservation Biology 15, 970–979.

Schneider, R., M. Humpert, K. Stoner, G. Steinauer, 2005. The Nebraska Natural Legacy

Project: A Comprehensive Wildlife Conservation Strategy. The Nebraska Game

and Parks Commission. Lincoln, Nebraska. August, 2005. 167 p.

Sinokrot, B.A., J.S. Gulliver, 2000. In-stream flow impact on river water temperatures.

Journal of Hydraulic Research 38, 339–350.

119

Smith, A.C., N. Koper, C.M. Francis, L. Fahrig, 2009. Confronting collinearity:

comparing methods for disentangling the effects of habitat loss and

fragmentation. Landscape Ecology 24, 1271–1285.

Smith, C.B., 2011. Adaptive management on the central Platte River – Science,

engineering, and decision analysis to assist in the recovery of four species. Journal

of Environmental Management 92, 1414–1419.

120

Supalla, R., K. Bettina, O. Yeboah, R. Bruins, 2002. A game theory approach to deciding

who will supply instream flow water. Journal of the American Water Resources

Association 38, 959–966.

Theberge, J.B., M.T. Theberge, J.A. Vucetich, P.C. Paquet, 2006. Pitfalls of applying

adaptive management to a wolf population in Algonquin Provincial Park, Ontario.

Environmental Management 37, 451–460.

Walters, C., 1997. Challenges in adaptive management of riparian and coastal

ecosystems. Conservation Ecology 1, 1. Available from:

http://www.consecol.org/vol1/iss2/art1 (Accessed March 2010).

Walters, C.J., 1986. Adaptive management of renewable resources. MacMillan, New

York, USA.

Williams, B. K., R.C. Szaro, C.D. Shapiro, 2007. Adaptive Management: The U.S.

Department of the Interior Technical Guide. Adaptive Management Working

Group, U.S. Department of the Interior, Washington, D.C.

Williams, B.K., 1996. Adaptive optimization and the harvest of biological populations.

Mathematical Biosciences 136, 1–20.

120

Zallen, M., 1997. Integrating New Values with Old Uses in the Relicensing of Kingsley

Dam and Related Facilities: Making Part of the Problem a Part of the Solution.

Presented at Dams, Water and Power in the New West, Eighteenth Annual

Summer Conference of the Natural Resources Law Center, School of Law

University of Colorado, Boulder.

121

Zuur, A.F., E.N. Ieno, G.M. Smith, 2007. Analysing Ecological Data. Springer Science +

Business Media, LLC, New York, USA.

121

122

Chapter 5: CONCLUSIONS

When decisions must be made regardless of the level of knowledge or uncertainty, the application of conventional research methods is often insufficient to support effective decision-making. There is a critical need to improve how research is incorporated into management decisions where uncertainty places limitations on contributions of science (Reynolds et al., 1996; Lee and Bradshaw, 1998; Berg et al.,

1999; Robertson and Hull, 2001). Adaptive management provides natural-resources managers with a method where they can learn through an ongoing process of implementing various management actions, monitoring management outcomes, and updating ecological models by comparing actual outcomes with expected outcomes by repeating decisions within an SDM approach (Hilborn and Walters, 1981; Walters, 1986;

Williams, 1996; Carpenter, 2002; Johnson et al., 2002). The application of adaptive management is appropriate for complicated natural-resources management problems with value-laden decisions high in risk and uncertainty (Levin, 1999; Gunderson and Holling,

2002; Berkes, 2004).

With multiple definitions of adaptive management available to natural-resources

122

122 managers, it was unclear which definitions were most useful and successful for managers.

In chapter 2, I reviewed adaptive management literature and assessed the level of success for two main schools of thought in adaptive management. Scientific literature acknowledges that successful application of adaptive management requires building a thorough understanding of the various elements of the process through cumulative experience (Gerber et al., 2007). If adaptive management is to improve as an approach to

123 management under uncertainty, it is imperative to study the process of adaptive management itself, including all approaches. Adaptive management as a concept continues to evolve through shifts in the dominant school of thought, as well as gain greater acceptance as a possible framework for management. The clearest recommendation from this review for increasing the implementation of AM is to deemphasize experimentation, taking a more passive AM approach.

In chapter 3, I presented a population model description focused on the Interior

Least Tern and Piping Plover populations on the Lower Platte River, Nebraska, and described how the model could be a vital aspect of an adaptive management plan. By utilizing the model description, managers gain a thorough understanding of model components and interactions between input and output parameters. The population model is a useful tool to aid resources managers in measuring the annual status of the two avian species and is best suited for a continuous process of reducing uncertainty through adaptive management. However, developing a clear adaptive management plan that utilizes the population model for the recovery of Least Terns and Piping Plovers will require collaboration among the multiple natural-resources agencies managing the LPR,

123

123 and especially the development of measurable objectives.

In chapter 4, I reported findings from a simulated multi-model analysis that assessed competing research hypotheses of an existing adaptive management plan, the

Platte River Recovery Implementation Program, with a multi-model inference framework. Prioritizing research hypotheses that focus on stronger parameter effect sizes will yield stronger model inference than weak effect sizes given limited monitoring

124 efforts. With thorough evaluation of research hypotheses and expected parameter effect sizes, the Program could maximize learning opportunities while proceeding with iterative decision-making. In maximizing learning, the Program will continue to decrease uncertainty about responses of Least Terns to various ecological factors on the central

Platte River (CPR), Nebraska, through adaptive management.

While managers in the field of natural resources generally acknowledge adaptive management as an appropriate approach for managing complicated ecosystems, the managers may experience difficulty in proceeding with the adaptive management process to the implementation stage. Increased efficacy in the adaptive management process may result from starting small (i.e., utilizing simple models) and incorporating structured decision-making. Overall, adaptive management provides a way to connect science to stakeholder values and allows managers to maximize learning while proceeding with necessary decision-making (Gregory and Keeney, 2002; Williams et al., 2007).

Additional increases in the efficacy of implementing an adaptive management plan include utilizing predictive models based on management objectives to increase learning. Placing less emphases on experimentation, allows managers to continue

124

124 through the adaptive management process while still learning about the managed system.

Predictive models that are flexible can be simply updated annually with monitoring data.

Developing models that are based directly on management objectives further increases the effectiveness of the models. Therefore, adaptive management plans should incorporate measurable objectives that allow managers to evaluate success.

125

Measurable objectives provide science an avenue to influence management effectively through utilizing objectives to develop various scenarios regarding expected and potential monitoring data. Managers could estimate how an unforeseeable event could influence the system and decrease the possibility of future surprises (Parma et al.,

1998; Greeuw et al., 2000). Development of possible extreme events through workshops with experts and stakeholders provides managers with a conceptual model of the effects of extreme events. By modifying model parameters to extreme values to represent unexpected events, the model output will aid in providing insight for resources managers on the inherent uncertainty of surprises.

125

126

LITERATURE CITED

Berg, J., B. Bradshaw, J. Carbone, C. Chojnacky, S. Conroy, D. Cleaves, R. Solomon, S.

Yonts-Shepherd, 1999. Decision Protocol 2.0. U.S. Forest Service FS-634.

Berkes, F., 2004. Rethinking community-based conservation. Conservation Biology 18,

621–630.

Carpenter, S.R., 2002. Building an ecology of the long now. Ecology 83: 2069–2083.

Gerber, L.R., J. Wielgus, E. Sala, 2007. A decision framework for the adaptive

management of an exploited species with implications for marine reserves.

Conservation Biology 21, 1594–1602.

Greeuw, S.C.H., M.B.A. van Asselt, J. Grosskurth, C.A.M.H. Storms, N. Rijkens-Klomp,

D.S. Rothman, J. Rotmans, 2000. Cloudy crystal balls. European Environ- ment

Agency, Copenhagen, Denmark.

Gregory, R.S., R.L. Keeney, 2002. Making Smarter Environmental Management

Decisions. Journal of the American Water Resources Association 38, 1601–1612.

Gunderson, L.H., C.S. Holling, (Eds.), 2002. Panarchy: understanding transformations in

human and natural systems. Island Press, Washington, D.C.

126

Hilborn, R., C.J. Walters, 1981. Pitfalls of Environmental Baseline and Process Studies.

EIA Review 2, 265–278.

Johnson, F A., W.L. Kendall, J.A. Dubovsky, 2002. Conditions and limitations on

learning in the adaptive management of mallard harvests. Wildlife Society

Bulletin 30, 176–185.

127

Lee, D.C., G.A. Bradshaw, 1998. Making monitoring work for managers. U.S. Forest

Service. http://www.icbemp.gov/spatial/lee_monitor/begin.html. Accessed 30

March 2010.

Levin, S. A., 1999. Fragile dominion: complexity and the commons. Perseus, New York.

Parma, A.M., NCEAS working group on population management, 1998. What can

adaptive management do for our fish, forests, food, and biodiversity? Integrative

Biology 1, 16–26.

Reynolds, K., P. Cunningham, L. Bednar, M. Sanders, M. Foster, R. Olson, D. Schmolt,

D. Latham, B. Miller, J. Steffenson, 1996. A knowledge-based information

management system for watershed analysis in the Pacific Northwest U.S. AI

Applications 2, 9–22.

Robertson, D.P., R.B. Hull, 2001. Beyond biology: toward a more public ecology for

conservation. Conservation Biology 15, 970–979.

Walters, C.J., 1986. Adaptive management of renewable resources. MacMillan, New

York, USA.

Williams, B. K., R.C. Szaro, C.D. Shapiro, 2007. Adaptive Management: The U.S.

127

Department of the Interior Technical Guide. Adaptive Management Working

Group, U.S. Department of the Interior, Washington, D.C.

Williams, B.K., 1996. Adaptive optimization and the harvest of biological populations.

Mathematical Biosciences 136, 1–20.

128

Appendix A: PEER-REVIEWED SCIENTIFIC LITERATURE USED IN THE

ANLAYSIS AND DISTRIBUTED BY SCHOOL OF THOUGHT

FOLLOWED BY SUCCESS CATEGORY

RESILIENCE EXPERIMENTALIST SCHOOL OF THOUGHT

Success Category: Theory

Berkes, F., J. Colding, C. Folke, 2000. Rediscovery of traditional ecological knowledge

as adaptive management. Ecological Applications 10, 1251–1262.

Berkes, F., 2004. Rethinking community-based conservation. Conservation Biology 18,

621–630.

Blumenthal, D., J. Jannink, 2000. A classification of collaborative management methods.

Conservation Ecology 4, 13 http://www.consecol.org/vol4/iss2/art13/. Accessed

30 March 2010.

Blumstein, D.T., 2007. Darwinian decision making: putting the adaptive into adaptive

management. Conservation Biology 21, 552–553.

Bradshaw, G. A., J.G. Borchers, 2000. Uncertainty as information: narrowing the

science–policy gap. Conservation Ecology 4, 7 128

128

http://www.consecol.org/vol4/iss1/art7/. Accessed 30 March 2010.

Keough, H.L., D.J. Blahna, 2006. Achieving integrative, collaborative ecosystem

management. Conservation Biology 20, 1373–1382.

Koontz, T.M., J. Bodine, 2008. Implementing ecosystem management in public agencies:

lessons from the U.S. Bureau of Land Management and the Forest Service.

Conservation Biology 22, 60–69.

129

Sutherland, W.J., 2006. Predicting the ecological consequences of environmental change:

a review of the methods. Journal of Applied Ecology 43, 599–616.

Wilhere, G.F., 2002. Adaptive management in habitat conservation plans. Conservation

Biology 16, 20–29.

Success Category: Suggest

Bax, N.J, R.E. Thresher, 2009. Ecological, behavioral, and genetic factors influencing the

recombinant control of invasive pests. Ecological Applications 19, 873–888.

Bennett, L.T., M.A. Adams, 2004. Assessment of ecological effects due to forest

harvesting: approaches and statistical issues. Journal of Applied Ecology 41, 585–

598.

Bierwagen, B.G., R. Thomas, A. Kane, 2008. Capacity of management plans for aquatic

invasive species to integrate climate change. Conservation Biology 22, 568–574.

Converse, S.J., G.C. White, K.L. Farris, S. Zack, 2006. Small mammals and forest fuel

reduction: National-scale responses to fire and fire surrogates. Ecological

Applications 16, 1717–1729.

Curtin, C., D. Western, 2008. Grasslands, people, and conservation: Over-the-horizon

129

learning exchanges between African and American pastoralists. Conservation

Biology 22, 870–877.

Doak, D.F., J.A. Estes, B.S. Halpern, U. Jacob, D.R. Lindberg, J. Lovvorn, D.H. Monson,

M.T. Tinker, T.M. Williams, J.T. Wootton, I. Carroll, M. Emmerson, F. Micheli,

M. Novak, 2008. Understanding and predicting ecology dynamics: are major

surprises inevitable? Ecology 89, 952–961.

130

Frederiksen, M., J.D. Lebreton, T. Bregnballe, 2001. The interplay between culling and

density-dependence in the great cormorant: a modeling approach. Journal of

Applied Ecology 38, 617–627.

Lindenmayer, D.B., C.R. Margules, D.B. Botkin, 2000. Indicators of biodiversity for

ecologically sustainable forest management. Conservation Biology 14, 941–950.

Mac Nally, R., E. Fleishman, D.D. Murphy, 2004. Influence of temporal scale of

sampling on detection of relationships between invasive plants and the diversity

patterns of plants and butterflies. Conservation Biology 18, 1525–1532.

Martin, K.L., L.K. Kirkman, 2009. Management of ecological thresholds to re-establish

disturbance-maintained herbaceous wetlands of the south-eastern USA. Journal of

Applied Ecology 46, 906–914.

McCarthy, M.A., K.M. Parris, 2008. Optimal marking of threatened species to balance

benefits of information with impacts of marking. Conservation Biology 22, 1506–

1512.

Redfield, G.W., 2000. Ecological research for aquatic science and environmental

restoration in south Florida. Ecological Applications 10, 990–1005.

130

Robinson, C.J., D. Smyth, P.J. Whitehead, 2005. Bush tucker, bush pets, and bush

threats: Cooperative management of feral animals in Australia's Kakadu National

Park. Conservation Biology 19, 1385–1391.

Sauer, J.R., M.G. Knutson, 2008. Objectives and metrics for wildlife monitoring. Journal

of Wildlife Management 72, 1663–1664.

131

Toledo, V. M., B. Ortiz-Espejel, L. Cortés, P. Moguel, M.D.J. Ordoñez, 2003. The

multiple use of tropical forests by indigenous peoples in Mexico: a case of

adaptive management. Conservation Ecology 7: 9

http://www.consecol.org/vol7/iss3/art9/. Accessed 30 March 2010.

Warburton, B., B.G. Norton, 2009. Towards a Knowledge-based ethic for lethal control

of nuisance wildlife. Journal of Wildlife Management 73, 158–164.

Wasserberg, G., E.E. Osnas, R.E. Rolley, M.D. Samuel, 2009. Host culling as an adaptive

management tool for chronic wasting disease in white-tailed deer: a modeling

study. Journal of Applied Ecology 46, 457–466.

Zweig, C.L., W.M. Kitchens, 2009. Multi-state succession in wetlands: a novel use of

state and transition models. Ecology 90, 1900–1909.

Success Category: Framework

Barrows, C.W., M.B. Swartz, W.L. Hodges, M.F. Allen, J.T. Rotenberry, B. Li, T.A.

Scott, X. Chen, 2005. A framework for monitoring multiple-species conservation

plans. Journal of Wildlife Management 69, 1333–1345.

Bearlin, A.R., E.S.G. Schreiber, S.J. Nicol, A.M. Starfield, C.R. Todd, 2002. Identifying

131

the weakest link: simulating adaptive management of the reintroduction of a

threatened fish. Canadian Journal of Fisheries and Aquatic Sciences 59, 1709–

1716.

Dimond, W.J., D.P. Armstrong, 2007. Adaptive harvesting of source populations for

translocation: a case study with New Zealand Robins. Conservation Biology 21,

114–124.

132

Gerber, L.R., J. Wielgus, E. Sala, 2007. A decision framework for the adaptive

management of an exploited species with implications for marine reserves.

Conservation Biology 21, 1594–1602.

Gray, A. N. 2000. Adaptive ecosystem management in the Pacific Northwest: a case

study from coastal Oregon. Conservation Ecology 4, 6

http://www.consecol.org/vol4/iss2/art6/. 30 March 2010.

Irwin, E.R., M.C. Freeman, 2002. Proposal for adaptive management to conserve biotic

integrity in a regulated segment of the Tallapoosa River, Alabama, U.S.A.

Conservation Biology 16, 1212–1222.

Jacobson, S.K., J.K. Morris, J.S. Sanders, E.N. Wiley, M. Brooks, R.E. Bennetts, H.F.

Percival, S. Marynowski, 2006. Understanding barriers to implementation of an

adaptive land management program. Conservation Biology 20, 1516–1527.

Long, J., A. Tecle, B. Burnette, 2003. Cultural foundations for ecological restoration on

the White Mountain Apache Reservation. Conservation Ecology 8, 4

http://www.consecol.org/vol8/iss1/art4/. Accessed 30 March 2010.

Lyons, J.E., M.C. Runge, H.P. Laskowski, W.L. Kendall, 2008. Monitoring in the context

132

of structured decision-making and adaptive management. Journal of Wildlife

Management 72, 1683–1692.

Marmorek, D., C. Peters, 2001. Finding a PATH toward scientific collaboration: insights

from the Columbia River Basin. Conservation Ecology 5, 8

http://www.consecol.org/vol5/iss2/art8/. Accessed 30 March 2010.

133

Moore, A.L., C.E. Hauser, M.A. McCarthy, 2008. How we value the future affects our

desire to learn. Ecological Applications 18, 1061–1069.

Richter, B.D., R. Mathews, D.L. Harrison, R. Wigington, 2003. Ecologically sustainable

water management: managing river flows for ecological integrity. Ecological

Applications 13, 206–224.

Salafsky, N., R. Margoluis, K.H. Redford, J.G. Robinson, 2002. Improving the practice

of conservation: a conceptual framework and research agenda for conservation

science. Conservation Biology 16, 1469–1479.

Walters, C., J. Korman, L.E. Stevens, B. Gold, 2000. Ecosystem modeling for evaluation

of adaptive management policies in the Grand Canyon. Conservation Ecology 4, 1

http://www.consecol.org/vol4/iss2/art1/. Accessed 30 March 2010.

Success Category: Implement

Gregory, R., D. Ohlson, J. Arvai, 2006. Deconstructing adaptive management: criteria for

applications to environmental management. Ecological Applications 16, 2411–

2425.

Hagmann, J.R., E. Chuma, K. Murwira, M. Connolly, P. Ficarelli, 2002. Success factors

133

in integrated natural resource management R&D: lessons from practice.

Conservation Ecology 5, 29 http://www.consecol.org/vol5/iss2/art29/. Accessed

30 March 2010.

Hansen, G.J.A., M.L. Jones, 2008. A rapid assessment approach to prioritizing streams

for control of Great Lakes sea lampreys (Petromyzon marinus): a case study in

134

adaptive management. Canadian Journal of Fisheries and Aquatic Sciences 65,

2471–2484.

Molina, R., B.G. Marcot, R. Lesher, 2006. Protecting rare, old-growth, forest-associated

species under the survey and manage program guidelines of the Northwest Forest

Plan. Conservation Biology 20, 306–318.

Shannon, J.P., D.W. Blinn, T. McKinney, E.P. Benenati, K.P. Wilson, C. O’Brien, 2001.

Aquatic food base response to the 1996 test flood below Glen Canyon Dam,

Colorado River, Arizona. Ecological Applications 11, 672–685.

Success Category: Mention

Beier, P., 2003. Adaptive management and SCB’s evaluation of species recovery plans.

Conservation Biology 17, 653–655.

Carpenter, S.R., 2002. Building an ecology of the long now. Ecology 83, 2069–2083.

Clark, J.A., J.M. Hoekstra, P.D. Boersma, P. Kareiva, 2003. Adaptive management and

SCB’s evaluation of species recovery plans. Conservation Biology 17, 655.

Stansfield, B., 2003. Adapative management and trial-and-error learning. Conservation

Ecology 7: 10 http://www.consecol.org/vol7/iss1/resp10/. Accessed 30 March

134

2010.

DECISION THEORETIC SCHOOL OF THOUGHT

Success Category: Theory

Dorazio, R.M., F.A. Johnson, 2003. Bayesian inference and decision theory – A

framework for decision making in natural resource management. Ecological

Applications 13, 556–563.

135

Runge, M.C., F.A. Johnson, 2002. The importance of functional form in optimal control

solutions of problems in population dynamics. Ecology 83, 1357–1371.

Success Category: Suggest

Buckley, Y.M., 2008. The role of research for integrated management of invasive

species, invaded landscapes and communities. Journal of Applied Ecology 45,

397–402.

Clarke, M.F., 2008. Catering for the needs of fauna in fire management: science or just

wishful thinking? Wildlife Research 35, 385–394.

Foster, S.J., A.C.J. Vincent, 2005. Enhancing sustainability of the international trade in

seahorses with a single minimum size limit. Conservation Biology 19, 1044–

1050.

Runge, M.C., J.R. Sauer, M.L. Avery, B.F. Blackwell, M.D. Koneff, 2009. Assessing

allowable take of migratory birds. Journal of Wildlife Management 73, 556–565.

Shea, K., D. Kelly, A.W. Sheppard, T.L. Woodburn, 2005. Context-dependent biological

control of an invasive thistle. Ecology 86, 3174–3181.

Wolf, N., M. Mangel, 2008. Multiple hypothesis testing and the declining-population

135

paradigm in Stellar sea lions. Ecological Applications 18, 1932–1955.

Success Category: Framework

Bogich, T., K. Shea, 2008. A state-dependent model for the optimal management of an

invasive metapopulation. Ecological Applications 18, 748–761.

136

Conroy, M.J., R.J. Barker, P.W. Dillingham, D. Fletcher, A.M. Gormley, I.M.

Westbrooke, 2008. Application of decision theory to conservation management:

recovery of Hector’s dolphin. Wildlife Research 35, 93–102.

Hauser, C.E., H.P. Possingham, 2008. Experimental or precautionary? Adaptive

management over a range of time horizons. Journal of Applied Ecology 45, 72–

81.

Martin, J., M.C. Runge, J.D. Nichols, B.C. Lubow, W.L. Kendall, 2009. Structured

decision making as a conceptual framework to identify thresholds for

conservation and management. Ecological Applications 19, 1079–1090.

McCarthy, M.A., H.P. Possingham, 2007. Active adaptive management for conservation.

Conservation Biology 21, 956–963.

Morellet, N., J. Gaillard, A.J.M. Hewison, P. Ballon, Y. Boscardin, P. Duncan, F. Klein,

D. Maillard, 2007. Indicators of ecological change: new tools for managing

populations of large herbivores. Journal of Applied Ecology 44, 634–643.

Rout, T.M., C.E. Hauser, H.P. Possingham, 2009. Optimal adaptive management for the

translocation of a threatened species. Ecological Applications 19, 515–526.

136

Shea, K., H.P. Possingham, W.W. Murdoch, R. Roush, 2002. Active adaptive

management in insect pest and weed control: intervention with a plan for learning.

Ecological Applications 12, 927–936.

137

Success Category: Implement

Armstrong, D.P., I. Castro, R. Griffiths, 2007. Using adaptive management to determine

requirements of re-introduced populations: the case of the New Zealand hihi.

Journal of Applied Ecology 44, 953–962.

Conn, P.B., W.L. Kendall, 2004. Evaluating Mallard adaptive management models with

time series. Journal of Wildlife Management 68, 1065–1081.

Cruz, F., V. Carrion, K.J. Campbell, C. Lavoie, C.J. Donlan, 2009. Bio-economics of

large-scale eradication of feral goats from Santiago Island, Galápagos. Journal of

Wildlife Management 73, 191–200.

Kilpatrick, A.M., C.M. Gillin, P. Daszak, 2009. Wildlife-livestock conflict: the risk of

pathogen transmission from bison to cattle outside Yellowstone National Park.

Journal of Applied Ecology 46, 476–485.

Linacre, N.A., A. Stewart-Oaten, M.A. Burgman, P.K. Ades, 2004. Incorporating

collateral data in conservation biology. Conservation Biology 18, 768–774.

Lynam, T., F. Bousquet, C. Le Page, P. d'Aquino, O. Barreteau, F. Chinembiri, B.

Mombeshora, 2002. Adapting science to adaptive managers: spidergrams, belief

137

models, and multi-agent systems modeling. Conservation Ecology 5: 24

http://www.consecol.org/vol5/iss2/art24/. Accessed 30 March 2010.

Shorey, R.I., K.T. Scribner, H.H. Prince, A.N. Kravchenko, D.R. Luukkonen, P.I.

Padding, 2007. Genetic analysis of standardized collections of Cackling and

Canada Goose harvests. Journal of Wildlife Management 71, 1458–1466.

138

Success Category: Mention

Mertens, S.K., J.M. Yearsley, F. van den Bosch, C.A. Gilligan, 2006. Transient

population dynamics in periodic matrix models: methodology and effects of

cyclic permutations. Ecology 87, 2338–2348.

SCHOOL OF THOUGHT: OTHER

Success Category: Theory

Clark, J.A., E. Harvey, 2002. Assessing multi-species recovery plans under the

Endangered Species Act. Ecological Applications 12, 655–662.

Hobbs, N.T., R. Hilborn, 2006. Alternatives to statistical hypothesis testing in ecology: a

guide to self teaching. Ecological Applications 16, 5–19.

Hulme, P.E., 2005. Adapting to climate change: is there scope for ecological

management in the face of a global threat? Journal of Applied Ecology 42, 784–

794.

Pandey, D.N., 2003. Cultural resources for conservation science. Conservation Biology

17, 633–635.

Parr, C.L., A.N. Andersen, 2006. Patch mosaic burning for biodiversity conservation: a

138

critique of the pyrodiversity paradigm. Conservation Biology 20, 1610–1619.

Stem, C., R. Margoluis, N. Salafsky, M. Brown, 2005. Monitoring and evaluation in

conservation: a review of trends and approaches. Conservation Biology 19, 295–

309.

139

Success Category: Suggest

Aburto-Oropeza, O., E. Sala, G. Paredes, A. Mendoza, E. Ballesteros, 2007.

Predictability of reef fish recruitment in a highly variable nursery habitat. Ecology

88, 2220–2228.

Bar-David, S., D. Saltz, T. Dayan, 2005. Predicting the spatial dynamics of a

reintroduced population: the Persian Fallow Deer. Ecological Applications 15,

1833–1846.

Cinner, J.E., T.R. McClanahan, N.A.J. Graham, M.S. Pratchett, S.K. Wilson, J. Raina,

2009. Gear-based fisheries management as a potential adaptive response to

climate change and coral mortality. Journal of Applied Ecology 46, 724–732.

Davis, G.E., 2005. Science and society: Marine reserve design for the California Channel

Islands. Conservation Biology 19, 1745–1751.

Goldingay, R.L., J.R. Stevens, 2009. Use of artificial tree hollows by Australian birds and

bats. Wildlife Research 36, 81–97.

Hirzel, A.H., B. Posse, P. Oggier, Y. Crettenand, C. Glenz, R. Arlettaz, 2004. Ecological

requirements of reintroduced species and the implications for release policy: the

139

case of the bearded vulture. Journal of Applied Ecology 41, 1103–1116.

Muller, B., A. Linstadter, K. Frank, M. Bollig, C. Wissel, 2007. Learning from local

knowledge: Modeling the pastoral-nomadic range management of the Himba,

Namibia. Ecological Applications 17, 1857–1875.

Olexa, E.M., P.J.P. Gogan, 2007. Spatial population structure of Yellowstone bison.

Journal of Wildlife Management 71, 1531–1538.

140

Rempel, L.L., M. Church, 2009. Physical and ecological response to disturbance by

gravel mining in a large alluvial river. Canadian Journal of Fisheries and Aquatic

Sciences 66, 52–71.

Robertson, B.C., N.J. Gemmell, 2004. Defining eradication units to control invasive

pests. Journal of Applied Ecology 41, 1042–1048.

Schmiegelow, F.K.A., D.P. Stepnisky, C.A. Stambaugh, M. Koivula, 2006. Reconciling

salvage logging of boreal forests with a natural-disturbance management model.

Conservation Biology 20, 971–983.

Stephens, S.L., L.W. Ruth, 2005. Federal forest-fire policy of the United States.

Ecological Applications 15, 532–542.

Theobald, D.M., T. Spies, J. Kline, B. Maxwell, N.T. Hobbs, V.H. Dale, 2005.

Ecological support for rural land-use planning. Ecological Applications 15, 1906–

1914.

Treves, A., K.U. Karanth, 2003. Human-carnivore conflict and perspectives on carnivore

management worldwide. Conservation Biology 17, 1491–1499.

Vanderwel, M.C., J.R. Malcolm, S.C. Mills, 2007. A meta-analysis of bird responses to

140

uniform partial harvesting across North America. Conservation Biology 21,

1230–1240.

Vester, H.F.M., D. Lawrence, J.R. Eastman, B.L. Turner II, S. Calme, R. Dickson, C.

Pozo, F. Sangermano, 2007. Land change in the southern Yucatan and Calakmul

Biosphere Reserve: Effects on habitat and biodiversity. Ecological Applications

17, 989–1003.

141

Success Category: Framework

Andow, D.A., A.R. Ives, 2002. Monitoring and adaptive resistance management.

Ecological Applications 12, 1378–1390.

Success Category: Implement

Cummings, J., N. Reid, I. Davies, C. Grant, 2005. Adaptive restoration of sand-mined

areas for biological conservation. Journal of Applied Ecology 42, 160–170.

AGAINST

Hinrichsen, R. A., 2000. Are there scientific criteria for putting short-term conservation

ahead of learning? No. Response to Kai N. Lee 1999: "Appraising Adaptive

Management". Conservation Ecology 4: 7

http://www.consecol.org/vol4/iss1/resp7/. Accessed 30 March 2010.

Jager, H.I., 2005. Genetic and demographic implications of aquaculture in white sturgeon

(Acipenser transmontanus) conservation. Canadian Journal of Fisheries and

Aquatic Sciences 62, 1733–1745.

Prato, T., 2005. Accounting for uncertainty in making species protection decisions.

Conservation Biology 19, 806–814.

141

142

Appendix B: EXTENDED AKAIKE’S INFORMATION CRITERION RESULTS

Table 1. Average Akaike’s Information Criterion (AIC) components, estimated frequencies of each model ranked as the best AIC model, and mean AIC model ranking over 100 iterations for each scenario and analysis method. Frequency AIC / ΔAICc / Mean Model Structure Kb w e as best QAICa ΔQAICd i Rank model No Effect

PAf

7* Rsandg * psandh 428.5 4 3.55 0.18 8 4 2 Rsand 428.4 2 3.40 0.15 8 5 6 Rsand + elevai + 429.0 4 4.00 0.12 5 5 vegcoj 10 Rsand + psand + 429.6 7 4.57 0.10 7 8 eleva + vegco + chwidk + pfishl 9 Pfish 430.7 2 5.75 0.10 14 7 1 Null model 430.6 1 5.63 0.07 16 5 8 Chwid 431.2 2 6.24 0.07 13 6 4 Eleva 431.1 2 6.08 0.07 11 5 5 Vegco 430.7 2 5.70 0.07 12 5 3 Psand 431.1 2 6.08 0.07 6 5 PQm 2 Rsand 244.2 2 1.55 0.19 5 4 1 Null model 244.4 1 1.79 0.16 51 3 9 Pfish 245.5 2 2.87 0.11 11 7 5 Vegco 245.5 2 2.85 0.10 10 5

142 7 Rsand * psand 246.4 4 3.80 0.09 4 5

142 3 Psand 245.7 2 3.07 0.09 4 5

8 Chwid 245.8 2 3.14 0.09 5 6 4 Eleva 245.7 2 3.06 0.09 8 5 6 Rsand + eleva + 246.6 4 4.04 0.07 2 5 vegco 10 Rsand + psand + 250.6 7 7.95 0.01 0 10 eleva + vegco + chwid + pfish n NA 2 Rsand 413.7 3 1.66 0.18 4 4

143

Table 1. continued. AIC / ΔAIC / Frequency Mean Model Structure K w QAIC ΔQAIC i Top Rank 1 Null model 414.2 2 2.11 0.14 38 4 9 Pfish 415.1 3 3.00 0.11 13 7 7 Rsand * psand 415.5 5 3.38 0.11 6 5 5 Vegco 415.1 3 3.01 0.09 13 5 3 Psand 415.3 3 3.23 0.09 6 5 8 Chwid 415.4 3 3.29 0.09 8 6 4 Eleva 415.3 3 3.21 0.09 9 5 6 Rsand + eleva + 415.7 5 3.64 0.08 3 5 vegco 10 Rsand + psand + 418.4 8 6.33 0.03 0 9 eleva + vegco + chwid + pfish Tapering Baseline PA 7 Rsand * psand 425.6 4 3.64 0.18 18 5 2 Rsand 425.6 2 3.64 0.18 30 6 10 Rsand + psand + 426.7 7 4.79 0.12 10 7 eleva + vegco + chwid + pfish 6 Rsand + eleva + 426.3 4 4.32 0.10 5 4 vegco 3 Psand 428.1 2 6.19 0.10 12 5 9 Pfish 429.4 2 7.46 0.08 6 6 8 Chwid 429.8 2 7.86 0.07 7 6

1 Null model 429.4 1 7.46 0.06 5 6

143

5 Vegco 429.7 2 7.79 0.05 4 6 143

4 Eleva 429.8 2 7.86 0.05 3 5

PQ 2 Rsand 244.9 2 1.72 0.23 39 4 1 Null model 246.1 1 2.88 0.13 21 4 3 Psand 246.4 2 3.16 0.13 10 5 7 Rsand * psand 247.1 4 3.84 0.10 6 6 9 Pfish 247.0 2 3.83 0.10 6 6 8 Chwid 247.3 2 4.10 0.08 7 6 4 Eleva 247.3 2 4.08 0.08 6 5

144

Table 1. continued. AIC / ΔAIC / Frequency Mean Model Structure K w QAIC ΔQAIC i Top Rank 5 Vegco 247.3 2 4.05 0.08 4 5 6 Rsand + eleva + 247.4 4 4.20 0.07 1 5 vegco 10 Rsand + psand + 251.2 7 7.98 0.02 0 9 eleva + vegco + chwid + pfish NA 2 Rsand 411.5 3 1.81 0.21 38 4 3 Psand 413.0 3 3.28 0.12 10 5

7 Rsand * psand 413.1 5 3.37 0.12 8 5 1 Null model 412.9 2 3.17 0.11 18 4 9 Pfish 413.6 3 3.96 0.09 6 7 6 Rsand + eleva + 413.5 5 3.80 0.08 3 5 vegco 8 Chwid 413.9 3 4.25 0.08 7 6 4 Eleva 413.9 3 4.22 0.08 6 5 5 Vegco 413.9 3 4.20 0.07 3 5 10 Rsand + psand + 416.0 8 6.30 0.04 1 9 eleva + vegco + chwid + pfish Similar Effects PA 7 Rsand * psand 425.3 4 3.95 0.20 23 5 10 Rsand + psand + 425.7 7 4.30 0.19 21 7 eleva + vegco + chwid + pfish 144

144 3 Psand 427.7 2 6.31 0.12 15 5 2 Rsand 427.0 2 5.60 0.10 10 5 6 Rsand + eleva + 427.1 4 5.73 0.09 7 5 vegco 8 Chwid 429.8 2 8.45 0.07 9 5 4 Eleva 429.7 2 8.28 0.07 5 5 1 Null model 429.7 1 8.36 0.06 6 6 9 Pfish 429.9 2 8.54 0.05 1 6 5 Vegco 430.4 2 8.97 0.04 3 5

145

Table 1. continued. AIC / ΔAIC / Frequency Mean Model Structure K w QAIC ΔQAIC i Top Rank PQ 2 Rsand 246.7 2 2.40 0.16 19 4 3 Psand 247.1 2 2.80 0.16 23 5 1 Null model 247.2 1 2.95 0.13 25 4 7 Rsand * psand 247.9 4 3.64 0.11 8 6 4 Eleva 248.2 2 3.89 0.09 7 5 8 Chwid 248.3 2 4.00 0.09 5 6 9 Pfish 248.3 2 4.02 0.08 7 7 6 Rsand + eleva + 248.9 4 4.59 0.07 3 5 vegco 5 Vegco 248.6 2 4.28 0.07 2 5 10 Rsand + psand + 251.6 7 7.31 0.03 1 9 eleva + vegco + chwid + pfish NA 2 Rsand 412.3 3 2.49 0.15 17 5 3 Psand 412.7 3 2.90 0.15 24 5 7 Rsand * psand 413.0 5 3.20 0.13 14 6 1 Null model 413.0 2 3.18 0.11 15 4 4 Eleva 413.8 3 3.98 0.09 7 5 6 Rsand + eleva + 414.0 5 4.19 0.08 5 5 vegco 8 Chwid 413.9 3 4.10 0.08 6 6

9 Pfish 413.9 3 4.12 0.08 6 6

5 Vegco 414.2 3 4.40 0.06 3 5 145

145

10 Rsand + psand + 415.4 8 5.62 0.06 3 8

eleva + vegco + chwid + pfish Strong Effects PA 10 Rsand + psand + 435.4 7 1.15 0.70 71 7 eleva + vegco + chwid + pfish 7 Rsand * psand 442.2 4 7.88 0.26 27 4 2 Rsand 461.8 2 27.50 0.02 1 7

146

Table 1. continued. AIC / ΔAIC / Frequency Mean Model Structure K w QAIC ΔQAIC i Top Rank 6 Rsand + eleva + 458.8 4 24.49 0.02 1 6 vegco 3 Psand 508.1 2 73.84 0.00 0 5 5 Vegco 522.8 2 88.49 0.00 0 3 1 Null model 527.6 1 93.37 0.00 0 9 8 Chwid 526.9 2 92.64 0.00 0 5 9 Pfish 527.8 2 93.51 0.00 0 6 4 Eleva 527.5 2 93.21 0.00 0 4 PQ 7 Rsand * psand 248.6 4 2.39 0.45 50 3 10 Rsand + psand + 248.6 7 2.36 0.42 37 7 eleva + vegco + chwid + pfish 2 Rsand 257.1 2 10.87 0.08 9 8 6 Rsand + eleva + 257.7 4 11.44 0.05 4 5 vegco 3 Psand 282.8 2 36.52 0.00 0 5

1 Null model 292.4 1 46.13 0.00 0 8

8 Chwid 293.0 2 46.76 0.00 0 5

5 Vegco 290.7 2 44.51 0.00 0 3 9 Pfish 293.4 2 47.21 0.00 0 6 4 Eleva 293.3 2 47.08 0.00 0 5 NA

10 Rsand + psand + 420.5 8 1.84 0.46 45 7

eleva + vegco + 146

chwid + pfish 146

7 Rsand * psand 420.9 5 2.29 0.43 45 4

2 Rsand 429.0 3 10.34 0.07 8 8 6 Rsand + eleva + 430.1 5 11.46 0.04 2 4 vegco 3 Psand 448.5 3 29.84 0.00 0 4 1 Null model 455.0 2 36.35 0.00 0 8 5 Vegco 454.6 3 35.94 0.00 0 3 8 Chwid 456.0 3 37.35 0.00 0 6 9 Pfish 456.2 3 37.61 0.00 0 6

147

Table 1. continued. AIC / ΔAIC / Frequency Mean Model Structure K w QAIC ΔQAIC i Top Rank Large Difference PA 4 Eleva 456.2 3 37.57 0.00 0 5 10 Rsand + psand + 429.6 7 2.66 0.41 44 7 eleva + vegco + chwid + pfish 7 Rsand * psand 430.9 4 3.98 0.41 44 4 6 Rsand + eleva + 437.8 4 10.87 0.09 6 6 vegco 2 Rsand 439.5 2 12.51 0.08 6 7 3 Psand 478.5 2 51.58 0.00 0 5 4 Eleva 493.5 2 66.55 0.00 0 5 8 Chwid 493.3 2 66.36 0.00 0 5 1 Null model 494.2 1 67.29 0.00 0 8 5 Vegco 494.0 2 67.05 0.00 0 3 9 Pfish 494.0 2 67.05 0.00 0 5 PQ 7 Rsand * psand 247.1 4 1.77 0.45 49 4 2 Rsand 249.7 2 4.43 0.25 31 7 10 Rsand + psand + 249.9 7 4.59 0.15 12 7 eleva + vegco + chwid + pfish 6 Rsand + eleva + 251.0 4 5.69 0.14 8 4 vegco

3 Psand 271.4 2 26.06 0.00 0 5

147

4 Eleva 279.9 2 34.56 0.00 0 7 147

1 Null model 279.2 1 33.94 0.00 0 6

8 Chwid 279.8 2 34.47 0.00 0 6 9 Pfish 280.1 2 34.79 0.00 0 6 5 Vegco 280.1 2 34.79 0.00 0 3 NA 7 Rsand * psand 415.2 5 1.70 0.43 46 4 2 Rsand 417.9 3 4.38 0.22 25 6 10 Rsand + psand + 417.1 8 3.63 0.22 18 7 eleva + vegco + chwid + pfish

148

Table 1. continued. AIC / ΔAIC / Frequency Mean Model Structure K w QAIC ΔQAIC i Top Rank 6 Rsand + eleva + 419.0 5 5.47 0.13 11 4 vegco 3 Psand 437.9 3 24.40 0.00 0 6 4 Eleva 444.2 3 30.71 0.00 0 7 1 Null model 443.3 2 29.78 0.00 0 6 8 Chwid 444.1 3 30.59 0.00 0 6 9 Pfish 444.3 3 30.84 0.00 0 6 5 Vegco 444.4 3 30.88 0.00 0 3 a-e QAIC = Quasi Akaike’s Information Criterion (AIC) for overdispersion; K = number of model parameters; ΔAIC = relative Akaike’s Information Criterion (AIC); ΔQAIC=relative Quasi-AIC; wi = Akaike weight f, m, n PA = Poisson distribution with AIC; PQ = Poisson distribution with Quasi-AIC; NA = Negative Binomial distribution with AIC g-l Rsand = bare river sand area; psand = bare sandpit area; eleva = bare sand elevation; vegco = percent vegetation cover; chwid = active channel width; pfish = number of prey fish *Number of each model as they appear in the model set

148