1. Searching for a good stocking policy for Lake Michigan salmonines Michael L. Jones and Iyob Tsehaye Quantitative Fisheries Center, Fisheries and Wildlife Michigan State University 1 Lake Michigan Decision Analysis - 2012

2. Decision Analysis Structured, formal method for comparing alternative management actions Main components:  Specify objectives  Identify management options  Assess knowledge and account for uncertainties  Use model to forecast possible outcomes Consider the possible consequences of a decision, rather than just predicting the most likely consequence 2 Lake Michigan Decision Analysis - 2012

3. The Big Question How many salmon and trout should we stock into Lake Michigan each year? more stocking leads to greater harvest, and thus benefits - unless… too much stocking leads to poor feeding conditions and increased mortality, but too little stocking may lead to negative effects of alewife on other species 3 Lake Michigan Decision Analysis - 2012

4. What we need to know… 1.How many salmon and trout are out there? 2.How much do they eat? 3.How capable are the prey fish of meeting this demand? 4.What happens to salmon and trout feeding (and survival) when prey numbers are low? 4 Lake Michigan Decision Analysis - 2012

5. Our approach 1. Analyze the past Salmonine abundance Salmonine consumption Prey fish production Supply vs demand 2. Forecast the future Simulation model Data we used Stocking and harvest Growth and diet data Prey fish survey data 5 Lake Michigan Decision Analysis - 2012

6. What does the past tell us? Lake Michigan Decision Analysis - 2012 6

7. How many salmon and trout are out there?  Total salmonine numbers have remained relatively stable since 1990  Reduced Chinook stocking has been offset by increased wild fish production  More recently, improved survival of older Chinook salmon has also offset reduced stocking 7 Lake Michigan Decision Analysis - 2012

8. How many salmon and trout are out there? Age-3 Chinook numbers Salmonine abundance 8 Lake Michigan Decision Analysis - 2012

9. How much do they eat?  Total consumption has remained fairly stable for last decade  Chinook salmon have accounted for more than half of total demand consistently since 1980  Large alewife accounted for more than 40% of total prey consumed since 1980, except in the late 1980s when small alewife dominated 9 Lake Michigan Decision Analysis - 2012

10. How much do they eat? Consumption by all species of salmon and trout 1 KT = 2.2 million lbs Consumption by predator type Consumption by prey type 10 Lake Michigan Decision Analysis - 2012

11. How capable are the prey fish of meeting this demand?  Predation rates on alewife have ranged from 25%-45% per year from the mid-1980’s to present  Predation mortality peaked in mid-1980’s and has approached peak levels again recently  Alewife (and rainbow smelt) recruitment is variable and not strongly related to adult abundance 11 Lake Michigan Decision Analysis - 2012

12. Index of total predation mortality on alewife How capable are the prey fish of meeting this demand? 12 Lake Michigan Decision Analysis - 2012

13. How capable are the prey fish of meeting this demand? Stock and recruitment

14. What happens to salmon and trout feeding when prey numbers are low? Chinook salmon consumption has declined when alewife abundance declined Similar, but weaker pattern for lake trout Lake Michigan Decision Analysis - 2012 14

15. What happens to salmon and trout feeding when prey numbers are low? 15 Lake Michigan Decision Analysis - 2012

16. What can we say about the future? Lake Michigan Decision Analysis - 2012 16

17. Policy simulation model Accounts for uncertainties: key uncertainties concern prey recruitment (supply) and predator feeding (demand) 17 Lake Michigan Decision Analysis - 2012 What we Know Management Decisions Prediction of Outcome LMDA

18. Multiple possible futures Alternative relationships that are consistent with the data

19.  The model forecasts possible future changes in fish populations and harvest, given a stocking policy There are many possible futures, so we need to look at the range of possible (likely) outcomes This range tells us what we think is most likely, but also what might happen Mainly we’re interested in how likely it is that bad things will happen Here’s how it works… Model results 19 Lake Michigan Decision Analysis - 2012

20. Generating results: First simulation Average biomass = 243 kT 20 Lake Michigan Decision Analysis - 2012

21. First simulation: average alewife biomass = 243 kt Generating results 21 Lake Michigan Decision Analysis - 2012

22. Generating results: Second simulation Average biomass = 52 kT 22 Lake Michigan Decision Analysis - 2012

23. Second simulation: average alewife biomass = 52 kt Generating results 23 Lake Michigan Decision Analysis - 2012

24. … and so on (e.g., results after 15 simulations) Generating results 24 Lake Michigan Decision Analysis - 2012

25. An example result: Status quo policy In 45 of 100 cases alewife biomass was between 100 and 500 kt: OK In 26 of 100 cases alewife biomass was less than 100 kt: BAD 25 Lake Michigan Decision Analysis - 2012