Ecopath Ecosim and fisheries management Realist Non

Ecopath, Ecosim, and fisheries management? ? “Realist” “Non believer” “Believer”

Some terminology ECOPATH - Builds a food web n ECOSIM - One way to make this web dynamic n ECOSPACE - An attempt at spatial modeling n n Ew. E (from www. ecopath. org) is one implementation. – Semi-black box – This useful for initial work (data exploration, basic tradeoffs, in other words, ecological priors) – Currently insufficient for extensive “formal confrontation” of models and data – Most work here was performed with Ecosim algorithms, outside of the black box

Model distribution n Ew. E (from www. ecopath. org) is one implementation. – Semi-black box – This is v. powerful for some uses (data exploration, tradeoffs) – Currently insufficient for extensive “formal confrontation” of models and data n Our aim is to build in both of these areas (current use and better tools).

In the beginning was DYNUMES. . . – ECOPATH – Began with Polovina 1984, updated by Christensen and Pauly (early 1990 s) - statistics added until current (year 2000) version. But basic equations are unchanged (and well-examined) for over 10 years. – ECOSIM (and ECOSPACE) – Recent work to make a food web dynamic, theory and practice new (some is un-reviewed with ad-hoc corrections). – Unified (open? ) format is strength

Why try to use the “whole” food web in a predictive model? n Eastern Bering Sea

What is the use of a mass-balance ecosystem model, anyway? n First and foremost, stock-scale (usually annual) data integration, hypothesis exploration.

Use may be a qualitative communication of trade-offs n n n This model may do whatever you want, there are good and bad examples. But what you have to do to get what you want may be very instructive. Don’t mistake the explorations for yield predictions.

Not a single species replacement: used in conjunction n First steps to “large marine” scale predator/prey management – – n Climate shifts vs. noise (pulses) vs. interspecies vs. fishing. The secret life of metrics (total system biomass? T. L. of catch? ) Ecological theory. Sensitive (but mysterious) species issues. Predator culling is a real issue – current back-of-the-envelope approaches may be worse (these models show culling may or may not work). But models are the only way. . . n Data quality – Reconciliation, sensitivity, and targeting new data. – Will more data help (the predator culling issue). n Radical re-design of “working” an ecosystem. – Command, control, or along for the ride?

Criticisms – It’s a model n So is everything

Criticisms – It’s a model on the wrong scale n Stocks, not processes

Criticisms – It’s a biomass dynamics model n It’s one tool among many.

Biomass dynamics are poor dynamics? ? ? n It’s our conceptual basis (MSY). n n Replace one set of assumption (constant Ms) with another (simplified age structure). “A balance between necessary complexity across species. ” A complement to age-structured (single or multi-species) models. The “why won’t our hypothesis work with simple models” challenge. n Where it breaks down, detail may be added (delay difference etc. ).

n Single species – – – n M(age) fixed, estimated Growth/Bio = fixed or DD at age Recruit(0) = f(N, B) Partial recruitment to fishery, spawning climate, other spp added through external parameters MSFOR – – – n Comparison (deterministic forecast) M(age) = f(pred(age), prey(age)) Growth/Bio fixed or DD at age Recruit(0) = f(N, B) Partial recruitment to fishery, spawning, ontogenetic (given data) Climate external Ecosim – – – One or two pools -2 pools have internal delay structure M(juv, adu) = f(Bpred, Bprey) Growth/Bio(juv/adu) = f(Bpred, B prey) Recruit(0) = f(Bpred, Bprey) (If one pool, recruit is fixed prop. Cost of growth) Knife-edged recruit to fishery, spawning, ontogenetic, climate still external Shapes capture some age structure (reduce parameters)

Criticisms – It’s an equilibrium biomass dynamics model n Mass-balance is a perturbable starting point – Mass-balance is not an equilibrium assumption. – (First, a look at the mass balance process). n In moving from Ecopath to Ecosim, an equilibrium is built. – This confrontation is the major work to discuss. – (Overcompensatory functional responses, etc. ).

A single trophic relationship [Q/B]j*DCij*Bj [P/B]j*Bj Bi Bj [P/B]j*Bj + Ii +Ei - Sj[Q/B]j*DCj*Bj - Other loss = 0 (Mass Balance)

Solving each unknown – [P/B]i*Bi*EEi = F*Bi [Sj[Q/B]j*DCj*Bj ]i n P/B, EE unknown: – top down (demand) solution. n Q/B unknown, B unknown: – top down / one prey item – Catch (F*B) should be known. – Diet composition must be known. – Generalized inverse for over- or under-determined models.

Sources of dissipation (Q/B) (P/B) (1 -G) (1 -EE)

Sources of dissipation (EE is the key). Q*G*EE (Q/B) (P/B) (1 -G) n (1 -EE) EE is what you don’t know about the system. – May include known time trends in the accounting (BA: biomass accumulation).

Mass-balance (Ecopath step) reconciles data - not in itself an equilibrium n Data issues always a mix of good, bad, and ugly n a different way of reconciling conflicts n n Combine and compare: Harvest/stock assessments n Diet data n Bioenergetics/growth n Mortality/rate studies n Lower trophic level production n

The (black) art of model balancing n n Benefit: you start to see the trade-offs (necessary correlations). It’s where you first address data quality. – Reconciliation of scales, techniques, and sources. – “What must you do to reconcile multiple single species assessments” n Like the black art of Bayesian priors

The equilibrium question n All models have an equilibrium. n Ecosim starts there: it’s an Ecopath to Ecosim transition issue. n Fast rebound (overcompensation) may be tuned. n Sensitivity approach may be implemented to fix this (spin up approach).

Bioenergetics Pop. Rates (Z is key) B P/B Q/B DC EE Catch BA etc. (mass accounting) M 2 GE M 0 Vul F (no B) Equilibrium built here, perturbed here Alternate stable states? ?

ECOPATH to ECOSIM n From a zero-dimensional equilibrium state to a zero-dimensional dynamic equation: Prey Q/B*Bj P/B*Bi Predator EE Prey c(Bi, Bj) Predator

Dynamics of overlap - (one predator one prey) “It’s cold down there!” V Bj B-V aij. Vij. Bj vij (Bi-Vij) Vij Bi - Vij vij. Vij d. Vij /dt = vij(Bi-Vij) - vij. Vij - aij. Vij. Bj Assume fast equilibrium for Vij

The appearance of Density Dependence n d. Vij /dt = vij(Bi-Vij) - vij. Vij - aij. Vij. Bj = 0 n Vij = vij. Bi/(2* vij + aij. Bj) n Cij (Bi, Bj) = aijvij. Bi. Bj (2* vij + aij. Bj) Prey biomass Cij (or Minstant) Cij /Bj Predator Biomass

Mathematically, halfway between the trickle and the vat n Cij (Bi, Bj) = vij. Bi ( 2 vij + 1 ) aij. Bj – Integrate limited smaller spatial and temporal dynamics (more or less) – Single “vulnerability” parameter X ~ 2 v/a. Bj ratio n AGE STRUCTURE: – Possible example: good evidence for this functional response, both by age (e. g. pollock) and by densitydependence (e. g. halibut).

One predator, many prey n Prey switching exists as a complex of 3 variables – base diet, vulnerability, feeding time to modify suitabilities n Captures some age-structure dynamics without the age structure – Basic assumption is that biomass is not independent of diet, age structure. n n Switch or die? Invasions/vast changes not captured. b a a a b b a b b

Age-structure simulation Smaller biomass implies younger age structure through changing relative vulnerability set by ‘v’ parameters.

MSFOR vs. Ecosim? n Different sides of the same coin Simplify age structure (Ecosim) or simplify consumption (MSVPA). n MSVPA assumes fixed suitabilities at age. n Ecosim assumes changing suitabilities with biomass (and therefore with age and foraging combined). n n Hybrid methods are quite possible.

Fishing in Ecosim n By individual species or gear type – may apply to a species directly, or as an effort multiplier to gear. n Gear type applies exploitation rate on multiple species group. . . bycatch is tied to gear effort.

Model behavior n Top-down (fishing) experiments: Apex predators behave as single-species models with (over? ) compensatory growth of prey. n Pella-Tomlinson form if prey is fixed. n Cascades appear below apex predators. n n Middle and lower trophic level fishing results are unpredictable.

MSY and overcompensation in base scenarios Phytoplankton n Zooplankton (Aydin 2001; submitted) Fish

The effect of vulnerability on MSY Phytoplankton Zooplankton Fish Catch Eq. Fish biomass (prop. of K) Zoop B/B 0 Eq. Fish biomass (prop. of K) Fish

age-structure and bioenergetics n Some basic decisions in the model need to be revisited in the next generation. – Coordination with MSVPA – Energy partitioning – Myers et al. – Bioenergetics decisions. – But led to compesation/depensation.

MSY and bioenergetic overcompensation n Another example: passive vs. active metabolism in zooplankton Phytoplankton Eq. Fish biomass (prop. of K) Zooplankton Fish

Fit to single species?

Additional data: anomalies in consumption – Systematic anomalies in consumption rates? n n n Food habits Predator size Prey size abundant year classes Age class models – Run the model backwards? Too much noise! – Evidence of alternate stable states?

Recruitment – A delay-difference equation with juveniles divided into monthly pools: n n n Size vs. age at recruitment tuneable Energy apportionment strategies Individual growth rates – Knife edge recruitment to fishery, spawning, and ontogenetic switch. – Spawning biomass is indirect measure. – This is a primary simplification (also for afternoon discussion).

Model behavior n Bottom-up (forcing) experiments: Time scale (frequency) is important. n Who responds the fastest? n Invasions are not predictable. n n Explanations may be dangerous External (climate) hypotheses must exist n (EBS climate fitting as case-study: afternoon) n climate image: n

mesoscale and migrations n Mesoscale Reasonable as single-species models for fishing experiments n Seasonal changes, aggregations on prey may lead to detectable systematic changes in foraging parameters n n Migrations n Model may be damped by “external” food sources.

Needed to make ECOSIM rigorous n Many of the problems listed (prey switching, etc. ) are not specific to Ecosim. – Basic fitting exists. – Thorough peer-reviewed testing against singlespecies, MSVPA models. – An improved statistical framework. – This is the next major development (come see the quantitative seminar!).

Fitting 1979 -2000 n First: – Vul fitting indicates low vuls (v<0. 05) fits better (recruitment dominated? ? ) – Kept vuls at 0. 3

The confrontation: can it be done, what do we learn? n Ecopath as priors. n Specification of full-scale problem in progress (balancing importance and covariance of bioenergetics, foraging, mortality).

Bioenergetics B P/B Q/B DC EE Catch BA etc. (mass accounting) Pop. Rates (Z is key) M 2 GE M 0 Vul F (no B) Fitting occurs here Ecopath as priors to examine correlation using: population and life history trade-offs, some single-species models

Meanwhile, culling in a simple model n Can we reasonable predict the results of a removal of the top predator? n Groundfish are near MSY: – – n F=M 80% of M from mammals 15% of M from pred. Fish 5% “unidentified” Can we increase yield (while holding effort constant) by removing mammals?

Yes (made to happen) (Truism: killing an animal will stop it from eating: but where does the energy end up? )

Our confidence? n Perform 1000 s of draws, allowing start out of equlibrium drawing: – Diets from uniform ± 30% – Vuls from range between 0. 1 and 0. 6 – All others (P/B, Q/B, passive/active respiration) from uniform ± 10%

Results: often down, not up Biomass after 50 years/start biomass n Mammal vs. predatory fish: fish wins – – – Improving diet data unlikely to help this picture. Possibility of improving mammals through lower fishing also uncertain. Admission: this is a simple, tightly-wired web (vuls tightly wired? ? ). n What about more complex webs? What about climate variability? What about the unmodeled, inedible predator? n WHAT ABOUT PROCESS UNCERTAINTY INCREASE? ? ? n n

Command, control, or along for the ride? n n n If we can’t predict manipulations (esp. in light of added climate variability), we should aim/add to our objectives the minimization of unpredictable cascades, rather than the optimization of multispecies yields or specific trophic-based rebuilding plans. A “healthy” ecosystem (without homeostasis). How much did all fish go through “regimes” before we fished them?

Laundry list 1 – Model savvy in current uses Endangered (im)possibilities (restore S. S. L. s through prey? ) n Climate and causality seeking. n Seeking key/critical species interactions and uncertainties. n Model behavior and improvement. n Radical rethinking (the impossible MSY and variability? ) n

Laundry List for management discussion (some may work or may have worked) First steps to “large marine” scale predator/prey management – – – n Necessary in a loosely/tightly wired system? Climate shifts vs. noise vs. interspecies vs. fishing. The secret life of metrics (total system biomass? T. L. of catch? ) Ecological theory (cultivation/depensation). Communication of alternatives. Endangered and HAPC species Issues. Predator culling is a real issue – current back-of-the-envelope approaches may be worse (these models show culling may or may not work). n Data quality issues – Reconciliation. – Sensitivity/targeting new data. n Radical re-design of “working” an ecosystem. – Command, control, or along for the ride?