1. 2. Multi-species models and concepts

2. No fish stocks live alone. Single species stock assessment is a simplification with important limitations: Technical interactions: How do we harvest on a single species basis? Biological interactions: Production is a result of biological interactions (food). What are the dynamics between harvesting and production on the ecosystem level?

3. Multispecies fisheries Technical interactions: What are the effects of direct mortality on target species and incidental mortality on other biota (bycatch)? Biological interactions: What are the effects of (fishery) changes in the flow of energy through the ecosystem? Single species models can partly answer the first, but multispecies models are need for the latter.

4. Fundamental drivers Ecological fundamentals: Most ecologists acknowledge two primary processes as the underlying forces governing populations: –Environmental disturbance (bottom up) –Trophodynamic interactions (top down): Competition, Predation (incl. fishing!) Life history strategies and parameters (growth, reproduction, survival etc.) are shaped from these.

5. Multispecies models Competition: Handled indirectly as change in abundance of one species is compensated in the trophic structure, by reciprocal changes in abundance of species with overlapping diets. Predation (incl. fishing!): Considered a main regulatory process. Multispecies models were initially aimed at estimating natural mortality as a function of predation. Environmental disturbance: Not handled. Most models are for retrospective and exploratory purposes. Few, if any, are used in management.

6. Environmental changes? What are the fishery, environmental, and trophic effects in historical data? How important are each of these? Catch records show decadal-scale variation in abundance synchronous with climate. Hardly any ‘standard’ models use physical forcing – simply because it is yet unpredictable. Recent attempts of ‘short-term’ predictions from multiple regression models. IPCC 2001, Box 6-1

7. North sea fisheries: Is it fishing or recruitment ?

8. Classification of multispecies models Handles lower trophic levels No Yes Handles age structure No Yes Handles age structure No Yes Multispecies Production models Age/size Structured MSVPA Aggregated ecosystem models ECOPATH Age/size structured individual based models Biological interactions Yes Technical interactions MSYPR No Start Modified from Hollowed et al. (2000)

9. Classification of multispecies models Handles lower trophic levels No Yes Handles age structure No Yes Handles age structure No Yes Multispecies Production models Age/size Structured MSVPA Aggregated ecosystem models ECOPATH Age/size structured individual based models Biological interactions Yes Technical interactions MSYPR No Start Modified from Hollowed et al. (2000)

10. Technical interactions: ‘Multispecies Y/R’ (MSYPR) Total (summed) Y/R for 4 species and 2 mesh sizes. Arrows correspond to max Y/R for each species, plus actual fishing effort. After Murawski (1984) Remember NO biological interactions! Larger mesh sizes = higher predation see MSVPA

11. Sum of single species production models for 3 species. NB this assumes same gear and q! A = overexploited, while C and B are under exploited MSY for C would deplete A Technical interactions: Surplus production models

12. Two species system: Krill–whales in Antarctica. Lotka-Volterra model. Krill yield will increase as whales are reduced and vice versa. MSY for each species is not sustainable whales will suffer. Biological interactions: Surplus production models Sustainable yields

13. SSVPA MSVPA Age 1 F M1M1 Age 2 F M1M1 Age 3 F M1M1 Age 1 F M1M1 Age 2 F M1M1 Age 3 F M1M1 Age 1 F M1M1 Age 2 F M1M1 Age 3 F M1M1 M2M2 M2M2 M2M2 M2M2 Prey Predator

14. Energy flow through a simple food web for the Northern California Current. Dark boxes and flows represent benthic energy paths and light boxes represent pelagic (new production) pathways; the size of the boxes is relative to the square root of the mean standing biomass. It can get very complicated

15. MSVPA Catch at age in numbers Terminal F’s Residual mortality (M 1 ) Predator consumption rates Body weight at age Predator stomach contents Abundance in numbers Predator consumption rates Body weight at age Predator stomach contents Abundance in numbers Body weight ‘MSVPA species’ ‘Other predators’ ‘Other prey’ Input Output Fishing mortality rates (F) Retrospective stock size (N) Suitability coefficients Predation mortality rates (M 2 ) ‘MSVPA species’ Reflect the preference of a prey to a predator = the diet composition of the predator relative to the available food

16. MSVPA species, other prey and other predators Biomass flow of the eastern Bering Sea system (arrows pointing to predators). After Jurado-Molina 2000.

17. MSVPA equations p = prey a = prey age i = predator j = predator age N = mean annual stock size R = annual ration W = weight of prey in stomach i S = suitability coefficient Up,a,i,j = fraction of prey p of age a found in stomach of predator i of age j These equations (1, 2 and 3) are solved iteratively within the MSVPA

18. What is the functional relationship between prey and predators? Behavioural response of predators to abundance of prey : Proportional to the occurrence? or Switching when prey become scarce? Take a guess

19. Stomachs, stomachs..stomachs 16389 stomachs of 5 species of predators cod, Whiting, Saithe, Mackerel and Haddock were sampled in 1981 in the North Sea (Daan 1987)

20. Diet compositions in the North sea Skjoldal (2004)

21. Annual consumptions Average annual biomass of prey species consumed by Atlantic cod in the North sea from 1974-1995. Cod predation to cod yield = 4:1 From ICES (1997).

22. Predation vs fishing Relative consumption of fish biomass from fish, mammals, birds and humans. After Bax 1991

23. Predation and fishing on cod Man is acting opposite all other predators!! How much does increased understanding of mortality on pre-recruit stages help us in management? From ICES (1997).

24. MSVPA Has it been used? Both Yes and No. –Results on M2 has been incorporated in SSVPA (M adjusted). –But MSVPA is not used in management Data requirements are insatiable – and is predation the main factor influencing processes?

25. Use of MSVPA Improved estimates on M (using M2 from MSVPA) in single species assessment. From ICES (1997).

26. North Sea multispecies system Effect of mesh change from 85 to 120 mm Percent changes in the long term fishery yields for North Sea stocks resulting from an increase in trawl mesh size from 85 to 120 mm for the directed fishery for cod. Results are presented for 1)single species (but multi-fleet) assessment. 2)MSVPA including interspecies predation Lower yields in the MSVPA results are due to greater predation rates from large predatory fish (cod, whiting, haddock, saithe) released by the larger mesh sizes. Source: Anonymous 1989. Report of the multispecies assessment working group. Int. Counc. Explor. Sea., C.M. 1989/Asess: 20, Copenhagen.

27. ECOPATH ( http://www.ecopath.org/ ) Trends in the number of publications citing ECOPATH in their abstract. From Dame & Christian 2006.

28. ECOPATH The Ecopath approach uses mass balance principles to estimate flows between user defined ecological groups. Each group is represented by one balanced equation and requires six input parameters: –biomass (B), –production (P), –consumption (Q), –ecotrophic efficiency (EE) –diet composition, and –catch of each group. The linear equations are solved via matrix algebra to produce estimates of the flows that balance inputs and outputs. Catch (export) and diet composition must always be given while of the four remaining basic input parameters (B, P, Q, and EE), three must be given and the fourth can be estimated. One free

29. ECOPATH ECOPATH is a static description (snapshot) of the ecosystem represented by biomasses aggregated into ecologically functional groups. Predators and prey are linked through consumption defined by a diet composition matrix and consumption rate parameters (Q/B). ECOPATH differs from traditional multi-species type models in that it does not require: –(i) representation of individual species, –(ii) age structure of species. The big difference between MSVPA and ECOPATH is that ECOPATH use direct data on total mortality rate (Z), in the form of the Production/Biomass ratio that users must provide. Production/Biomass = Z when mortality can be modeled as a constant negative exponential function (Allen 1971)

30. Defining ecological groups http://www.cmrh.dk/HMAP/Ainsworthhistoricalmodels.pdf

31. Input equals output: Consumption(Q) = production(P) - unassimilated food - respiration: Production = predation + other mortality + fishery The basic assumptions EE i = ecotrophic efficiency, EX i = export, B i = biomass, PB i = production/biomass ratio, QB i = consumption/biomass ratio DC ij = fraction of prey i in diet of predator j Group QP Production (P) = Consumption (Q) – U - R Production (P) = Consumption (Q) ∙ Growth efficiency (GE)  GE also called ‘Food conversion factor’ in aquaculture 1 kg fish = 4 kg food: GE = ¼ = 0.2 (20%)

32. Other mortality Harvest Consumption Predation Other mortality Other mortality Predation Respi- ration Harvest Unassimilated food Respi- ration Unassimilated food Unassimilated food Respi- ration Mass balance: sharing the pie

33. Data requirements for Ecopath For each group, provide estimates in green, and the program will estimate those in red. Choose one: 1). B, P/B, Q/B, EE, DCs,... 2). B, P/B, Q/B, EE, DCs,... 3). B, P/B, Q/B, EE, DCs,... 4). B, P/B, Q/B, EE, DCs,... Ranked ease of estimation: P/B and Q/B > B > DCs >> EE hence EE often left unknown (Option 1).

34. Ecological efficiencies ecotrophic efficiency (EE) –The proportion of the production of a trophic level taken by consumers.trophic levelconsumers growth efficiency (GE) –The weight increment of the consumers divided by the weight of food consumed.consumers ecological efficiency (TE) –The amount of energy produced from a given trophic level divided by that supplied to the same trophic level. This is the product of the ecotrophic efficiency and the growth efficiencytrophic levelecotrophic efficiencygrowth efficiency

35. Parametres (B and P/B) Biomasses are obtained from standard assessment methodologies Production/Biomass ratio:  From catch composition data using standard stock assessment methodologies;  Natural mortality of fish from Pauly’s (1980) empirical equation:  M = K0.65 · Loo-0.279 ·T0.463  F = catch / biomass;  P/B = Z = F + M.

36. Q/B (Food consumption) - The tail story The faster swimming fish eats more

37. Food consumption - The tail story Q/B = 3 · W  -0.2 · T 0.6 · A R 0.5 · 3 e Ft W  = asymptotic weight T = temperature A R = aspect ratio = F t = foodtype (0 f. carn.) Aspect ratio: Yellow Red A R = 9.8 A R = 1.3 Height 2

38. (DC) Diet compositions e.g., for a tuna Use volume or weight! Partly digested fish 31.6% Others 19.3% Portunids 15.8% Euphausiids 3.5% Squids 12.3% Anchovies 8.8% Sardines 7% Auxids 1.7%

39. Output flows from ECOPATH Flow diagram of the central South China Sea pelagic ecosystem in the 1980s. Arrows indicate flow (t km -2 y -1 ) and boxes (size ≈ log 10 of B) the biomass (t km -2 ). Walters et al. (1997) How do we use this in management?

40. Used to explore ecosystem form and functioning A wide variety of analyses are developed in theoretical ecology (key journal: Ecological Modeling). Many are included in Ecopath  ecotrophic efficiency (EE)  growth efficiency (GE)  ecological transfer efficiency (TE) Network analyses from ECOPATH

41. EcoSim and EcoSpace Builds on steady state, stock-recruitment relationships and constant environment Plaice Sole Cod Haddock Saithe Whiting Sandeel Herring Sinclair et al. (2004)

42. Multispecies vs single species If fisheries management is to: Conduct a historical reconstruction of the stock to establish key parameters and relationships and to describe the current stock status (assessment). Propose specific actions (e.g. total allowable catch) that will steer the stock towards a desired status (short-term forecasting). Make long-term predictions of the likely future status of the stock under various management scenarios to establish desirable states (long-term forecasting). Advise on the robustness of management procedures (precautionary approach). Then it is questionable whether multispecies models provide better advice than single-species models. Hollowed at al (2000)

43. Multispecies vs single species Estimates of recruitment based on single-species and multispecies models from six ecosystems (EBS: eastern Bering Sea; GOA: Gulf of Alaska). Temporal trends in predation mortality account for only a minor fraction of the interannual variability in marine production. Hollowed at al (2000)

44. The optimum fishing pattern is non-selective ! How much ? How ? Is fishing down the food web bad? Where is the highest production? ? ? Why make it complicated ? What if we fish everything proportionally? Fishing every component proportional to their M preserves the relative balance of the ecosystem. The system remains unchanged, except everything is less. This means fishing everything!

45. The virgin ‘community’ B1 B2 B3 MSY Multi-species à la Kolding (1994 )

46. B1 B2 MSY MSY/2 B2 Fish the community from the top MSY MSY/2 1.Take MSY from B3 2.This will relieve predation from B2 3.So we can take MSY/2 from B2 Kolding (1994 )

47. B1 B2 MSY MSY/2 Fish the community from the bottom MSY MSY/2 1.Take half MSY from B1- leave other half for B2! 2.This will reduce carrying capacity for B2 and cascade up 3.Continue fish up the food web MSY/2 B2 Kolding (1994 )

48. What have we learnt? The sum of the whole is less than the sum of the parts! System models suggest that that the overall potential yield from a multispecies fishery is likely to be less than the sum of the individual potential yields. Long-term predictions from single-species and multispecies models remain uncertain because of the potential influence of environmental variation on recruitment Hollowed at al (2000)

49. EAF by 2010? What are the fishery, environmental, and trophic effects in historical data? IPCC 2001, Box 6-1

50. Concluding remarks Models can account only for factors contemplated by their authors. Models create scenarios and belief systems that may or may not correspond to reality. They can influence human behaviour for better or worse. Richard et al. 2004