Modeling And Multiobjective Risk Decision Tools for Ecosystem Management

1. Notice: The views expressed here are those of the individual authors and may not necessarily reflect the views and policies of the United States Environmental Protection Agency (EPA). Scientists in EPA have prepared the EPA sections, and those sections have been reviewed in accordance with EPA’s peer and administrative review policies and approved for presentation and publication. The EPA contributed funding to the construction of this website but is not responsible for it's contents. Mention of trade names or commercial products does not constitute endorsement or recommendation for use.

2. Modeling And Multiobjective Risk Decision Tools for Ecosystem Management Benjamin F. Hobbs1 Richard Anderson2 Jong Bum Kim1 Joseph F. Koonce3 1. Dept. of Geography & Environmental Engineering, The Johns Hopkins University 2. National Oceanic & Atmospheric Administration 3. Dept. of Biology,Case Western Reserve University

3. Goal of Presentation • Demonstrate how ecosystem-based fisheries management can be joined with ecological risk analysis under multiple management objectives

4. Goal of Presentation • Demonstrate how ecosystem-based fisheries management can be joined with ecological risk analysis under multiple management objectives • Introduce tools: • Ecosystem model • Multiobjective tradeoff analysis • Bayesian evaluation of ecological research

5. I. Unresolved Problems in Lake Erie • Major decline of fisheries in 1990s • Unknown effects of exotic species • Zebra Mussel invasion since 1988 • Round Goby increase in 1990s • Expected invasion of Ruffe

6. I. Unresolved Problems in Lake Erie • Major decline of fisheries in 1990s • Unknown effects of exotic species • Zebra Mussel invasion since 1988 • Round Goby increase in 1990s • Expected invasion of Ruffe • Declining productivity caused by decrease in P loading • Uncertain role of habitat

7. Historical Variation in Fish Harvest and Environment in Lake Erie Harvest Trends

8. Historical Variation in Fish Harvest and Environment in Lake Erie Harvest Trends Phosphorus Loading

9. II. Modeling Tradeoffs among Productivity, Exotics, & Fisheries: Lake Erie Ecological Model (LEEM) PO 4 Zooplankton Edible Algae Pred. Prey Fish Fish Inedible Algae Zoobenthos

10. II. Modeling Tradeoffs among Productivity, Exotics, & Fisheries: Lake Erie Ecological Model (LEEM) PO 4 Zooplankton Edible Algae Zebra Pred. Prey Mussel Fish Fish Inedible Algae Zoobenthos

11. II. Modeling Tradeoffs among Productivity, Exotics, & Fisheries: Lake Erie Ecological Model (LEEM) Stocking PO Fishing Mortality, Habitat 4 Zooplankton Edible Algae Zebra Pred. Prey Mussel Fish Fish Inedible Algae Zoobenthos Harvest

12. II. Modeling Tradeoffs among Productivity, Exotics, & Fisheries: Lake Erie Ecological Model (LEEM) Stocking PO Fishing Mortality, Habitat 4 Zooplankton Edible Algae Zebra Pred. Prey PCB Mussel Fish Fish Inedible Algae Zoobenthos Harvest

13. The Need for Multispecies Management: Effects of Species Interactions on Max Sustained Yield • Optimal exploitation of predator varies with fishing rates of prey species

14. The Need for Multispecies Management: Effects of Species Interactions on Max Sustained Yield • Optimal exploitation of predator varies with fishing rates of prey species

15. Interaction of Walleye Harvest & Phosphorus Loading Walleye abundance/harvest has a greater influence on total fish biomass than P loading

16. Total Fish Biomass 6.0 5.0 4.0 3.0 2.0 1.0 0.0 W5 W4 P5 P4 W3 P3 W2 P2 W1 Walleye Effort P Loading Interaction of Walleye Harvest & Phosphorus Loading Walleye abundance/harvest has a greater influence on total fish biomass than P loading

17. Implications of LEEM Studies • Fisheries and P Loading Jointly Determine Optimal Exploitation of Species

18. Implications of LEEM Studies • Fisheries and P Loading Jointly Determine Optimal Exploitation of Species • Derivation of Quotas for Single Species without Considering Interactions Can Lead to Overexploitation • Prey and predators cannot be managed independently

19. III. Value of Research: Multiobjective Bayesian Framework • Information has value only if it can change decisions and improve outcomes • “Value” is multidimensional!

20. III. Value of Research: Multiobjective Bayesian Framework • Information has value only if it can change decisions and improve outcomes • “Value” is multidimensional! • Necessary elements: 1. Management context: Alternatives, objectives, decision rule

21. III. Value of Research: Multiobjective Bayesian Framework • Information has value only if it can change decisions and improve outcomes • “Value” is multidimensional! • Necessary elements: 1. Management context: Alternatives, objectives, decision rule 2. What we know now: “States of nature” (hypotheses, parameter distributions), and confidence in each (“prior probabilities”)

22. III. Value of Research: Multiobjective Bayesian Framework • Information has value only if it can change decisions and improve outcomes • “Value” is multidimensional! • Necessary elements: 1. Management context: Alternatives, objectives, decision rule 2. What we know now: “States of nature” (hypotheses, parameter distributions), and confidence in each (“prior probabilities”) 3. Research:Information options (monitoring, modeling, experiments), and what might be learned

23. III. Value of Research: Multiobjective Bayesian Framework • Information has value only if it can change decisions and improve outcomes • “Value” is multidimensional! • Necessary elements: 1. Management context: Alternatives, objectives, decision rule 2. What we know now: “States of nature” (hypotheses, parameter distributions), and confidence in each (“prior probabilities”) 3. Research:Information options (monitoring, modeling, experiments), and what might be learned 4. System response: LEEM, expert judgment

24. III. Value of Research: Multiobjective Bayesian Framework • Information has value only if it can change decisions and improve outcomes • “Value” is multidimensional! • Necessary elements: 1. Management context: Alternatives, objectives, decision rule 2. What we know now: “States of nature” (hypotheses, parameter distributions), and confidence in each (“prior probabilities”) 3. Research:Information options (monitoring, modeling, experiments), and what might be learned 4. System response: LEEM, expert judgment 5.Integrating framework: A way of determining how information affects our knowledge and choices: Decision trees, Bayes’ rule

25. Problem Structure • Two decision stages • Research project; eh • P loading and fisheries • management;as ={as1, as2, as3, as4}

26. Problem Structure • Two decision stagesUncertainties θ • Research project; eh –Lower trophic level; • P loading and fisheries – Other uncertainties; • management;as ={as1, as2, as3, as4}

27. Problem Structure • Two decision stagesUncertainties θ Outcomes • Research project; eh –Lower trophic level; – 10 Attributes X, combined • P loading and fisheries – Other uncertainties; using additive utility • management;as ={as1, as2, as3, as4} function U(X)

28. θ) P( Problem Structure • Two decision stagesUncertainties θ Outcomes • Research project; eh –Lower trophic level; – 10 Attributes X, combined • P loading and fisheries – Other uncertainties; using additive utility • management;as ={as1, as2, as3, as4} function U(X) U(X(as, θ)) as No Research ,e0 P(θ |Z1) Research, e1 as P(Z1) Research, eh P(Zh) Research Decision E Outcomes from Research, eh Decision A Chance node for θ Utility Outcome

29. Management Levers • Phosphorus Loads: • 5K, 10K (as is), and 15K ton/yr

30. Management Levers • Phosphorus Loads: • 5K, 10K (as is), and 15K ton/yr • Exploitation effort: A measure of thenumber of boats or the time they spend fishing • Exploitation: Trawl, Gill Nets, and Sport Harvest • Base = historical exploitation level • Vary exploitation by + 50%

31. Present State of Knowledge • Prior probabilities • Uncertainties in LEEM parameters

32. Present State of Knowledge • Prior probabilities • Uncertainties in LEEM parameters • Hypotheses presented at 1999 IAGLR Modeling Summit and Lake Erie Millenium Conference • Changes in structure of lower trophic level (e.g., Zoobenthos production efficiency ) • The role of zebra mussels in Lake Erie energy and nutrient flows (e.g.,Zebra mussel recycling nutrients; Primary productivity as function of P loading)

33. Present State of Knowledge • Prior probabilities • Uncertainties in LEEM parameters • Hypotheses presented at 1999 IAGLR Modeling Summit and Lake Erie Millenium Conference • Changes in structure of lower trophic level (e.g., Zoobenthos production efficiency ) • The role of zebra mussels in Lake Erie energy and nutrient flows (e.g.,Zebra mussel recycling nutrients; Primary productivity as function of P loading) • Disregarding uncertainties may result in inappropriate, nonrobust decisions

34. Options for Gathering Information • Characteristics of research • Cost & time • Reliability of outcomes

35. Options for Gathering Information • Characteristics of research • Cost & time • Reliability of outcomes • Estimating the value of research • Research revises prior probabilities  “Posterior probabilities” (new state of knowledge) • New knowledge may influence management decisions • Calculate value by simulating decisions with and without new information

36. Ecosystem Health & Human Well-being Ecological Economic Social Recreation Consumption Self-sustaining Economically important species = Self- sustaining Native species Productivity Structure Function Multiple Objective Framework for Risk Analysis

37. Ecosystem Health & Human Well-being Ecological Economic Social Recreation Consumption Self-sustaining Economically important species X1: density of walleye X3: PCB conc in small mouth bass X8: commercial walleye biomass X9: commercial yellow perch biomass X10: commercial smelt biomass X2: annual mean walleye sport X7: Biomass ratio native Self- sustaining Native species species to total Productivity Structure Function X4: total biomass X5: walleye-percid X6: piscivore-planktivore biomass ratio biomass ratio Multiple Objective Framework for Risk Analysis

38. Bayesian Analysis Results • Priorities for objectives can affect decisions • All participants prefer High trawling; most High sport harvest • Split on Gill net effort

39. Bayesian Analysis Results • Priorities for objectives can affect decisions • All participants prefer High trawling; most High sport harvest • Split on Gill net effort • Ignoring uncertainty can change decisions • True for 2 of 6 participants • Uncertainty about Zoobenthos productivity effects of zebra mussels most important (perfect information changes decisions) • Uncertainties about Zooplankton productivity and Zebra mussel recycling also important

40. Bayesian Analysis Results • Priorities for objectives can affect decisions • All participants prefer High trawling; most High sport harvest • Split on Gill net effort • Ignoring uncertainty can change decisions • True for 2 of 6 participants • Uncertainty about Zoobenthos productivity effects of zebra mussels most important (perfect information changes decisions) • Uncertainties about Zooplankton productivity and Zebra mussel recycling also important • The value of research stems (in part) from its effect on decisions. Research has value for 5 of 6 participants • Two projects most valuable: • Goby predation on mussels • Lakewide estimates of productivity • Worth: 101 - 104 tons/yr equivalent of Walleye sport harvest

41. Summary • Heuristic application of LEEM can lead to multi-fishery rules that recognize uncertainties (P, invasions, habitat)

42. Summary • Heuristic application of LEEM can lead to multi-fishery rules that recognize uncertainties (P, invasions, habitat) • “Ecosystem Health” can be operationalized E.g., Lake Erie stakeholders compared alternative futures using fuzzy cognitive maps and multiobjective analysis. Value judgments combined diverse “health” attributes, such as productivity, aesthetics, & community structure.

43. Summary • Heuristic application of LEEM can lead to multi-fishery rules that recognize uncertainties (P, invasions, habitat) • “Ecosystem Health” can be operationalized E.g., Lake Erie stakeholders compared alternative futures using fuzzy cognitive maps and multiobjective analysis. Value judgments combined diverse “health” attributes, such as productivity, aesthetics, & community structure. • Multiobjective Bayesian analysis can include ecological uncertainties in management, and quantify the value of research E.g., fish managers made value and probability judgments for a risk analysis, & showed that intensive monitoring of lower trophic level productivity could improve fisheries management

44. Take Home Message: • Methods to model the decision-making process itself (multiobjective tradeoff analysis, decision trees, Bayesian risk analysis) provide an important complement to science intended to develop indicators of ecosystem health • Could be applied to MAIA or any region to support ecosystem management

45. Acknowledgments • Research support provided by the International Joint Commission and US Environmental Protection Agency (STAR R82-5150) • US and Canadian environmental & natural resources managers and stakeholders for participating in modeling workshops and providing data and guidance in model development