Same data different story:

Transcription

1 Same data different story: guidelines for data weighting in a multispecies statistical catch-at-age stock assessment framework Kelli F. Johnson and André E. Punt School of Aquatic and Fishery Sciences, University of Washington, Seattle, WA October 21, 2015 Johnson & Punt Multispecies data weighting 1 / 20

2 1 Why 2 Multispecies model 3 Case study 4 Simulation 5 Next steps Johnson & Punt Multispecies data weighting 2 / 20

3 Why perform stock assessments Goal of stock assessment models:... understand, and inform decision-makers of, the consequences of possible fishing activities. (Hollowed et al., 2000) Johnson & Punt Multispecies data weighting 3 / 20

4 Why perform stock assessments Goal of stock assessment models:... understand, and inform decision-makers of, the consequences of possible fishing activities. (Hollowed et al., 2000) Ecosystem based management... shorthand for more holistic approaches to resource management. (Larkin, 1996) Johnson & Punt Multispecies data weighting 3 / 20

5 Why perform stock assessments Goal of stock assessment models:... understand, and inform decision-makers of, the consequences of possible fishing activities. (Hollowed et al., 2000) Ecosystem based management... shorthand for more holistic approaches to resource management. (Larkin, 1996) Goal of multispecies stock assessment models: explicitly represent species interactions, providing a framework for evaluating ecosystem properties and improved estimates of management quantities. Johnson & Punt Multispecies data weighting 3 / 20

6 Barriers to multispecies stock assessments increased data requirements Johnson & Punt Multispecies data weighting 4 / 20

7 Barriers to multispecies stock assessments increased data requirements increased uncertainty in model output Johnson & Punt Multispecies data weighting 4 / 20

8 Barriers to multispecies stock assessments increased data requirements increased uncertainty in model output decreased transparency associated with increasing complexity Johnson & Punt Multispecies data weighting 4 / 20

9 Barriers to multispecies stock assessments increased data requirements increased uncertainty in model output decreased transparency associated with increasing complexity lack methods for calculating management reference points Johnson & Punt Multispecies data weighting 4 / 20

10 Barriers to multispecies stock assessments increased data requirements increased uncertainty in model output decreased transparency associated with increasing complexity lack methods for calculating management reference points Johnson & Punt Multispecies data weighting 4 / 20

11 Statistical catch-at-age multispecies models age-structured forward projection

12 Statistical catch-at-age multispecies models age-structured forward projection maximum likelihood

13 Statistical catch-at-age multispecies models age-structured forward projection maximum likelihood

14 Statistical catch-at-age multispecies models age-structured forward projection maximum likelihood

15 Statistical catch-at-age multispecies models age-structured forward projection maximum likelihood Assessment Method for Alaska (AMAK; J. Ianelli)

16 Statistical catch-at-age multispecies models age-structured forward projection maximum likelihood Assessment Method for Alaska (AMAK; J. Ianelli) Kinzey and Punt, 2009

17 Statistical catch-at-age multispecies models age-structured forward projection maximum likelihood Assessment Method for Alaska (AMAK; J. Ianelli) Kinzey and Punt, 2009 Van Kirk et al., 2010, 2012, 2015

18 Statistical catch-at-age multispecies models age-structured forward projection maximum likelihood Assessment Method for Alaska (AMAK; J. Ianelli) Kinzey and Punt, 2009 Van Kirk et al., 2010, 2012, 2015 Curti et al., 2013

19 Statistical catch-at-age multispecies models age-structured forward projection maximum likelihood Assessment Method for Alaska (AMAK; J. Ianelli) Kinzey and Punt, 2009 Van Kirk et al., 2010, 2012, 2015 Curti et al., 2013 Johnson et al., 201? Johnson & Punt Multispecies data weighting 5 / 20

20 Case study Aleutian Islands, Alaska Johnson & Punt Multispecies data weighting 6 / 20

21 Case study Aleutian Islands, Alaska walleye pollock (Theragra chalcogramma) Johnson & Punt Multispecies data weighting 6 / 20

22 Case study Aleutian Islands, Alaska walleye pollock (Theragra chalcogramma) Atka mackerel (Pleurogrammus monopterygius) Johnson & Punt Multispecies data weighting 6 / 20

23 Case study Aleutian Islands, Alaska walleye pollock (Theragra chalcogramma) Atka mackerel (Pleurogrammus monopterygius) Pacific cod (Gadus macrocephalus) Johnson & Punt Multispecies data weighting 6 / 20

24 Case study Aleutian Islands, Alaska walleye pollock (Theragra chalcogramma) Atka mackerel (Pleurogrammus monopterygius) Pacific cod (Gadus macrocephalus) foodweb (blue = predator) Johnson & Punt Multispecies data weighting 6 / 20

25 Fixed inputs Pacific cod Atka mackerel walleye pollock Johnson & Punt Multispecies data weighting 7 / 20

26 Fixed inputs Pacific cod Atka mackerel walleye pollock Steepness: fixed at individual assessment values (0.7, 0.8, & 1.0) Johnson & Punt Multispecies data weighting 7 / 20

27 Fixed inputs Pacific cod Atka mackerel walleye pollock Steepness: fixed at individual assessment values (0.7, 0.8, & 1.0) Johnson & Punt Multispecies data weighting 7 / 20

28 Data moderate system Johnson & Punt Multispecies data weighting 8 / 20

29 Operating model Johnson & Punt Multispecies data weighting 9 / 20

30 Methods Weighting distribution weight types normal se year survey lognormal cv catch multinomial n age, length, & diet comps Johnson & Punt Multispecies data weighting 10 / 20

31 Methods Weighting distribution weight types normal se year survey lognormal cv catch multinomial n age, length, & diet comps Data source type OM EM 1 EM 2 EM 3 EM 4 fishery catch 0.05 fishery age & length survey index survey age & length survey diet weight survey diet length Johnson & Punt Multispecies data weighting 10 / 20

32 Model performance metrics Unfished spawning biomass Unfished recruitment Biomass available to survey Spawning biomass Annual recruitment Fishing mortality Johnson & Punt Multispecies data weighting 11 / 20

33 Results Johnson & Punt Multispecies data weighting 12 / 20

34 Results Johnson & Punt Multispecies data weighting 13 / 20

35 Results Johnson & Punt Multispecies data weighting 14 / 20

36 Results Johnson & Punt Multispecies data weighting 15 / 20

37 Results Johnson & Punt Multispecies data weighting 16 / 20

38 Results Johnson & Punt Multispecies data weighting 17 / 20

39 Results Johnson & Punt Multispecies data weighting 18 / 20

40 Next steps Coding tasks Johnson & Punt Multispecies data weighting 19 / 20

41 Next steps Coding tasks Initial conditions Johnson & Punt Multispecies data weighting 19 / 20

42 Next steps Coding tasks Initial conditions Lognormal survey likelihood Johnson & Punt Multispecies data weighting 19 / 20

43 Next steps Coding tasks Initial conditions Lognormal survey likelihood Iterative reweighting (McAllister & Ianelli, 1997; Stewart & Hamel, 2014) Johnson & Punt Multispecies data weighting 19 / 20

44 Next steps Coding tasks Initial conditions Lognormal survey likelihood Iterative reweighting (McAllister & Ianelli, 1997; Stewart & Hamel, 2014) Add 2004:2015 data Johnson & Punt Multispecies data weighting 19 / 20

45 Next steps Coding tasks Initial conditions Lognormal survey likelihood Iterative reweighting (McAllister & Ianelli, 1997; Stewart & Hamel, 2014) Add 2004:2015 data Add OMs and EMs Johnson & Punt Multispecies data weighting 19 / 20

46 Next steps Coding tasks Initial conditions Lognormal survey likelihood Iterative reweighting (McAllister & Ianelli, 1997; Stewart & Hamel, 2014) Add 2004:2015 data Add OMs and EMs Add process error Johnson & Punt Multispecies data weighting 19 / 20

47 Next steps Coding tasks Initial conditions Lognormal survey likelihood Iterative reweighting (McAllister & Ianelli, 1997; Stewart & Hamel, 2014) Add 2004:2015 data Add OMs and EMs Add process error Weighting methods Johnson & Punt Multispecies data weighting 19 / 20

48 Next steps Coding tasks Initial conditions Lognormal survey likelihood Iterative reweighting (McAllister & Ianelli, 1997; Stewart & Hamel, 2014) Add 2004:2015 data Add OMs and EMs Add process error Weighting methods Tasks for others or post-doc Johnson & Punt Multispecies data weighting 19 / 20

49 Next steps Coding tasks Initial conditions Lognormal survey likelihood Iterative reweighting (McAllister & Ianelli, 1997; Stewart & Hamel, 2014) Add 2004:2015 data Add OMs and EMs Add process error Weighting methods Tasks for others or post-doc Two species model Johnson & Punt Multispecies data weighting 19 / 20

50 Next steps Coding tasks Initial conditions Lognormal survey likelihood Iterative reweighting (McAllister & Ianelli, 1997; Stewart & Hamel, 2014) Add 2004:2015 data Add OMs and EMs Add process error Weighting methods Tasks for others or post-doc Two species model Move beyond self-test and estimate using Atlantis data Johnson & Punt Multispecies data weighting 19 / 20

51 Acknowledgements Dr. Doug Kinzey (SWFSC) Alaska Fisheries Science Center Northwest Fisheries Science Center NOAA / Sea Grant Population Dynamics Fellowship Johnson & Punt Multispecies data weighting 20 / 20