Table of Contents

_{Quick link to: EFIMAS home, WP3 }

A series of workshops on the economic aspects of Management Strategy Evaluation in relation to EFIMAS has been held during the project. The main aim of the workshops has been to provide information for the economic and bio-economic methodologies and models being developed for FLEcon, i.e. the economic package of the FLR framework inside the EFIMAS and COMMIT projects. The findings of these 'ECONOWS' workshops are collected in the Final report ECONOWS. This document presents a compilation of the existing and new approaches presented by economists at EFIMAS ECONOWS workshops held in Seville 2004, Copenhagen 2007 and 2008. Further results from EFIMAS project meetings in London and Nantes 2006 are included.

One of the main goals of EFIMAS has been to provide a framework in which different bio-economic methodologies (models) can be implemented and used next to one another in order to evaluate a variety of management measures. The economic methodology is designed to fit biological methodologies used for fish stock projections rather than focusing on con-ventional economic modelling concepts only in which fishermen’s behaviour and the economic incentive structure plays an important role. For the economic part of the framework this means:

- Specifying the methods used in the different models in use

- Building a framework that can cope with the different methods

- Implementing the main methods within the framework

The report mentioned above is structured along these lines. Section 2 of the report gives a summary of economic concepts. Section 3 describes the manner in which these concepts have been implemented in existing and new (bio-)models. The section ends with a summary of the initial specifications of key model equations included in FLEcon. Section 4 addresses the structure of FLEcon and its most important classes of tools. More comprehensive descriptions of existing models and approaches are attached in the appendices of the report.

Below is given a short outline of the economic models and modules developed specifically in connection with EFIMAS.

Within the ECONOWS workshop, parts of the FLEcon package have been created. This package contains the common functional forms of price dynamic models and production functions, as well as the Fcube model, the AHF model and methods for computing standard economic indices.

The Dynamic Capacity Change Model, also known as the ‘AHF model’, evaluates the dynamic change in fleet capacity from one time period to the next, given expectations about future earnings from the fishery. The central equation of the Capacity Change model is the evaluation of fleet capacity V (number of vessels) in year y for fleet segment b:

V_{y,b}= V_{y-1,b}+_{y,b} ; V_{MIN,b} ≤ V_{y,b} ≤ V_{MAX,b}

V_{y,b} = I_{IN,b} ∙ Π_{y,b} / p_{IN,b} ; Π_{y,b} ≥ 0

V_{y,b} = I_{OUT,b} ∙ Π_{y,b} / p_{OUT,b} ; Π_{y,b} < 0

Π_{y,b} is the capitalisation of future payments, that reflects the assumption that the fishermen base their investment/disinvestment decisions on expected future earnings. I_{IN,b} and I_{OUT,b} are the fractions of positive relatively negative expected profits that are used to invest/disinvest. p_{IN,b} and p_{OUT,b} are the prices per unit capacity of investment/disinvestment. It is assumed that the change in capacity is determined by the opportunity cost of capital including an option for asymmetry in entry and exit. The price of a vessel p_{IN,b} transforms pecuniary capital into physical capital, and the reciprocal of p_{OUT,b} includes the fisherman’s perception of opportunity costs. For more details on the investment/disinvestment function, refer to Hoff and Frost (2006).

The investment/disinvestment equation leading to dynamic capacity change presented above, is only one part of the overall bio-economic feedback model constructed to evaluate a chosen management scheme. The dynamic capacity change module has at present been used in three different settings:

- EFIMAS CS2: Demersal roundfish fisheries in the North Sea.
- Danish Seiners catching Cod in the North Sea.
- Dutch Beam Trawlers catching Plaice and Sole in the North Sea.

In the two latter cases special emphasis has been put on modelling the combined quota and effort (sea days) control imposed in many European fisheries in the North Sea. The models put special emphasis on that harvest (quota) and effort are necessarily interrelated, i.e. that a given harvest (quota) taken will necessarily determine the effort used, and correspondingly, that a given effort limit imposed will determine the harvest taken. Thus one of the two regulations will always be the limiting factor in a combined quota-effort regulation system. The two last bio-economic models shown above both take this casual relationship between harvest and effort into account, and as such switch between quota and effort control depending on which is the limiting factor.

See also:

Final report ECONOWS, Section 3.5 and Appendix 4.

Model Code: Dutch Beam Trawl Example,Danish Seine Example

The Fcube method (Ulrich et al., in prep) was developed in relation to EFIMAS Case Study 2 (North Sea Roundfish), as well as ICES SGMixMan (Study Group on Mixed Fisheries Management) and FP6 project AFRAME. Its objective is to explore and evaluate the potential overquota catches arising from inconsistent single-species TAC, based on simple assumptions about fleets effort distribution. Its strength is the relative simplicity of data needs while being able to account for a great diversity of fleets, metiers and stocks. It is running with FLR 2.0.

**Algorithm**:

Data are structured by fleet segment Fl (e.g. group of vessels), metier Mt (type of activity practiced by the fleet; a fleet can engage in several metiers over a year) and stock St. Catchability estimates by fleet and metier q, as well as effort distribution by fleet and metier Effshare in the TAC year Y is user-input, e.g. as historical average or use of alternative models. Then average catchability by fleet is estimated as

q(Fl,St,Y)=∑_{Mt}(q(Fl,Mt,St,Y)*Effshare(Fl,Mt,Y))

The fishing mortality corresponding to the single-stock TAC (Ftarg) is converted into “Stock dependent fleet effort”. The “stock-dependent fleet effort” is the estimated effort a fleet should develop in order to catch its quota share for a particular stock. The total target fishing mortality Ftarget(St) is first divided across fleet segments (partial fishing mortalities) through assumptions about quota share (e.g. historical landing share or alternative model) These partial fishing mortalities are subsequently used for estimating the stock-dependent fleet effort:

F(Fl,St,Y)=Ftarg(St,Y)*QuotaShare(Fl,St)

E(Fl,St,Y)=F(Fl,St,Y)/q(Fl,St,Y)

It is unlikely that the effort corresponding to each single-species TAC is the same across species, and the resulting effort is therefore a choice. The user can explore the outcomes of a number of assumptions or rules about fleets own behaviour (e.g. going on fishing after some quotas are exhausted) or management scenarios (e.g. all fisheries are stopped when the quota of a particular stock is reached).

E(Fl,Y)=rule(E(Fl,St1,Y),E(Fl,St2,Y),E(Fl,St3,Y)…)

Final effort by fleet is then used to recalculate the actual fishing mortality by stock and corresponding catches. Difference between catches and TAC is interpreted as overquota.

See also:

The methodology incorporates fisher behaviour in a full feed back model for mixed fisheries. The methodology introduces a simple and practical algorithm for a nonlinear catch - input relationship based on accepted theory in economics (decreasing returns to effort). The basic assumptions is that when subject to effort restrictions fishermen will skip those trips from which they expect the lowest earnings per unit of effort. The methodology indicates a convex catch – input relationship in the case of effort management in contrast to a concave relationship when applied to TAC driven scenarios. The catchability was defined as:

q_{t}=q_{0}(E_{t}/E_{0})^{β}

Where q = catchabilty and E = Effort and β is an estimated parameter. Estimation of the parameter β is based on cross sectional analysis of variation in results by vessel and by trip. Data were drawn from logbook data by vessel and by trip, prices by month and by species from landings statistics. The catchability equation and other economic equations fit into stock dynamics models prepared by IMARES and based on standard procedures applied by ICES. The model is applied for NS flatfish fisheries but is applicable for other EU fleet segments where stock assessment data and landings by trip are available.

See also:

In this exercise we are trying to evaluate the Northern hake long term management alternatives presented in (SEC (2007)). It implies that is not a pure economic evaluation of the situation of the fleet, given that we are restricting the evaluation to the effort drivers presented in it.

With this in mind, we present an evaluation using FLR (Kell et al., 2007) and its economic module FLEcon for pursuing advances in the economic evaluation of management strategies.

We have focused our work on two trawl fleets, that capture around the 30% of the total northern hake stock, with the characteristic that one (pair trawlers) face almost a single target fishery (hake is in average the 88% of the total landings) while the other is facing a multi-species fishery (where hake accounts for 64% of their landings), even if they first face hake as the target species.

This last characteristic is extremely important since the evaluation of the management strategy is based on several assumptions that are based on a single species framework. The most important one and the driver of the whole system is that Hake’s TAC is defining the effort applied, that is, the other species will be subrogated to this effort (even if in some alternative cases this assumption can be relaxed). In terms of conditioning we have read this assumption as dividing species types of production function in three categories (explained in the conditioning section). Firstly hake is considered management driven in terms that all TAC is always caught, and that the corresponding effort is calculated assuming a fixed catchability per age. Given that effort, and through the estimation, when availability of data makes it possible, of a production function for each fleet and species (anglerfish, and megrim), catches of the rest of the species are calculated. Finally, a third group of species is explicitly considered but catches are calculated based on an ad hoc calibration. The overall captures are also calibrated using what we have called (probably, in not a very fortunate way), “ghost species”.

A full explanation of the aplication made can be found in: MSE for Northern Hake.

**Fleet Dynamics**

The long term fleet based behavioural model used to simulate the entry-exit from the northern hake fishery depends on three related decisions: a) investment; b) decommissioning; c) selling in the second hand market. These decisions are simulated at each time t during the simulation period and each of them determines the change in the operating number of vessels.
Decisions are based on the average value of vessels. Four average values are considered: 1) The profit value estimated by the sum of actualized profits expected along the lifespan of a vessel (V_{Π}); 2) The value of a decommissioned vessel which has been estimated as the decommissioning rate (€/GT) times the average GT per vessel (V_{DG}); 3) The market value of an existing vessel in the second hand market estimated as the average price per GT paid by the market times the average GT per vessel (V_{SH}); 4) the average investment value of a new vessel estimated as the average price of a new GT times the average GT per vessel (V^{IN}).

a) Investment: Investment (I^{+}) is simulated only if the profit value is positive, V_{}Π>0. Thus,(I^{+}) is an increasing function of the profit value. Its slope (0.000001145) has been estimated based on time series regression analysis. Thus, the number of new vessels can approximated by equation (1).

ΔN´_{t+1}= I_{t}^{+} Π_{t}/V_{t}^{IN} (1)

b) Decommissioning: Since obviously the decommissioning value cannot be negative, the maximum number of vessels to be decommissioned (ΔN_{DG}) is calculated as the ratio of the decommissioning grant (DG_{t}) for the fleet and the decommissioning value (V_{DG,t}) per vessel. The number of vessels abandoning the fleet (ΔN´´_{t+1}) (2) is estimated as a linear function of the difference between the profit value (V_{DG,t}) and the decommissioning value. The estimated coefficients for β and γ are 4.10 and 0.00000469 respectively.

ΔN´´_{t+1}= β-γ(V_{Π,t} - V_{DG,t}) (2)

c) Selling: The number of vessels potentially sold on the market (ΔN´´´_{t+1})(3) is a function of the difference between the profit value (V_{Π}) and the market value (V_{SH}). The higher this difference, the lower the number of vessels sold on the market. The estimated β' and γ' coefficients are respectively 0.3422 and 0.00000088.

ΔN´´´_{t+1}= β'-γ'(V_{Π,t} - V_{SH,t}) (3)

Based on a), b) and c) the simulated number of vessels at time t+1 is defined by equation 4.

ΔN_{t+1}= ΔN´_{t+1} - ΔN´´_{t+1} - ΔN´´´_{t+1} (4)

Notice that since the number of vessels is an integer value, the partial changes due to investment, decommissioning and selling in the second hand market are approximated to the closer integer.

The link between fleet and effort dynamics
In many fisheries the fishing effort represents the main management control variable. In our case the fishing effort, estimated in terms of days at sea for each fleet, allows measuring the impact of the simulated variation in the number of vessels on the biological and economic indicators. It is defined by equation (5), where ds_{t} is days at sea and ds_{t} N_{t} is the number of vessels at time t.

E_{t} = ds_{t}/N_{t} (5)

R code: fleet_dynamics.zip

A dynamic state variable model can be used to evaluate the optimal strategy for a fishing vessel under annual individual landing or effort quota. Dynamic state variable models assume that optimal fishing behaviour can be calculated under the assumption that each individual is a utility maximizer. Although many other incentives may play a role in fishermen's behaviour, there is some empirical evidence for profit as the metric of utility. Dynamic state variable models allow one to mix the timescales between choices and constraints like fishermen facing an annual quota system but making daily, weekly ormonthly decisions on where to fish and which fish to keep on board. The individual vessels in the model may be constrained by the quota for the individual species and will respond by changing their fishing pattern in terms of (i) the number of fishing trips; (ii) the choice of fishing areas; (iii) the choice to discard the over-quota part of their catch.

The problem for the individual is thus to optimize utility function Φ, in this case the net revenue at the end of the year T, The utility function is based on total landings for the two species L_{1} and L_{2}, total fishing effort and travel time E and their respective prices p_{1}, p_{2} and variable costs p_{e}, taking into account the fine D for exceeding the individual quota

Φ(L_{1} ,L_{2} ,E) = L_{1} p_{1} + L_{2} p_{2} - D - E p .

The fine is calculated as a function of the quota q_{1}, q_{2} and the landings, such that the fine is 0 when none of the quota are exceeded and become larger as the quota exceed the landings. Now the expected revenue at the end of the year needs to be linked to the choices in the preceding
time steps. This is done using the value function, which is the maximum expected revenue between month t and the end of the year T, expressed as F(L_{1},L_{2},E,t). At the end of the year it is by definition equal to the utility function (F(L_{1},L_{2},E,T)= Φ). In the months preceding T, the function depends on the expected revenue consequences V_{ijk} (L_{1},L_{2},E,t) of visiting an area i and discarding an excess of j tonnes of the catch of species 1 and discarding an excess of k tonnes of the catch of species 2:

V_{ijk}(L_{1} ,L_{2}, E,t) =∑λ_{ij}(l_{1} ,t) ∑λ_{ik}(l_{2} ,t)F(L_{1} +l_{1} ,L_{2}+ l_{2} ,E+ e_{i} ,t+1)

In this equation, λ_{ij}(l_{1},t) is the probability that a vessel during a specific time step (month) will land l_{1} tonnes of fish of species 1, given a visit to area i and a choice to discard everything in excess of j.

Likewise, λ_{ij}(l_{2},t) is the probability that a vessel during a specific time step will land l_{2} tonnes of fish of species 2, given a visit to area i and a choice to discard everything in excess of k. The parameter e_{i} is the amount of effort, needed to visit area i.
The probabilities λ_{ij}(l_{1},t) and λ_{kj}(l_{2},t) can be seen as a two stage process. First, the probability of a catch is calculated using discretized normal distributions with mean μ_{1it}, μ_{2it} and standard deviation σ_{1}, σ_{2} for the two respective species. Then, the probability for the actual landings is adjusted by assuming that the probability of a landings equal to j, respectively k is equal to the probability of an excess catch. All landings in excess of the discarding choice are set to zero.

Finally, the stochastic dynamic programming equation that provides the optimal choice for the areas to visit and the discarding of marketable fish is

F(L_{1}, L_{2}, E, t)= max_{ijk}{V_{ijk}(L_{1}, L_{2}, E, t)},

and calculated in a backward iteration, because of the recursive nature of as F(L_{1},L_{2},E,t), which depends on V_{ijk} (L_{1},L_{2},E,t), in turn depending on F(L'_{1},L'_{2},E',t+1). The optimal fishing strategy is an array H(L_{1},L_{2},E,t) defining the optimal fishing strategy with respect to i,j and k in each timestep, given the state variables L_{1}, L_{2}, E. After the backward iteration, forward iterations can be done for a number of individuals, who choose the optimal path, defined by the optimal strategy, given the stochastic nature of the catches.

See also:

Test assumptions and scenarios to anticipate fishing effort redistribution effect on the evolution of stocks appears to be a key point to design and ensure the efficiency of regulations integrating the fishermen behaviour in the exercise. The modeling of the spatio-temporal fleet dynamic in response to fluctuations of the ressources in space and time and to the implementation of regulations (e.g. MPA, effort reduction) requires a particular spatially- and temporally- explicit model frame. Two models were used in parallel to investigate these topics. First, the existing ISIS-Fish software for fisheries modeling (Mahevas & Pelletier 2004) was used. Additionally, a new simulation frame was developed in R plugged to the FLR platform. Both models consists of three sub-model components: (i) a multi-species population module supporting spatial demographic processes and co-occurence between populations, (ii) a multi-fleet module taking into account of the heterogeneity of the fishing practices and (iii) a management module applying conventional management (e.g. TAC, effort control) or spatio-temporal closure. All these components operate on a spatial grid matching the resolution of underlying data at a monthly time step. The fleets dynamics are derived from log-book data and are possibly controlled via dynamic fishermen behaviour following pre-specified equations in response to implemented management regimes and/or stock fluctuations. Both models also includes an economic description of the fishery. Economic conditions may impact as feedback mechanism the fishing effort displacement in space and/or the reallocation between activities via change in fishermen behaviour. fba-model-efimasbrussels-10march08.pdf The stock-specific area- and season- age- disaggregated fishing mortalities is computed from the partial F obtained for each fishing activity:

partialF_{(fleet, gear, species, area, season)}= selectivity_{(species, gear)} x Effort_{(fleet, gear, area, season)} x efficiency_{(species, fleet, gear)} x catchability _{(species, -area-, -season-)}

F_{(species,area,season)}= Σ_{fleet} Σ_{gear} partialF_{(fleet, gear, species, area, season)}

and applying them on the stock abundances, makes the link with the stock evolutions through the classic survival equation. Total catches from the array of species are then dispatched between fleets according to their contribution to the total Fs. Stock-specific fish prices and (spatialized) costs for fishing enable the economic revenue from landings to be computed at the metier level (variable costs i.e. depending on effort at sea) and summed at the fleet level (sum of revenues over metiers minus fixed costs). A particular emphasize is put in the FLR model to break the total fishing effort into key allocations playing at different levels to act as entry points to test re-allocation:

Effort_{(country,vessel.category,fleet,gear,season,area)}= totNbVessels x propInCountry x propInVesselCategory x propInFleet x propInActivity x effort.per.vessel x propInGear x propInArea

ISIS-Fish is first dedicated to compare the effect on stock and fleet evolutions of various designs of spatio-temporal closures across the fishing activities. Then, a particular stress using this model are put for testing closure and effort reallocation side effect. Using the FLR model on the other side, different management regimes are compared in a management procedure framework: (i) a one-year lag TAC system in which a TAC to reach a F_{target} is set each year+1 from the result of the conventional short-term forecast based on the XSA assessment of the population at y-1; (ii) a one-year lag effort control management in which total allowed effort for the year+1 is decided from the F_{target}, (iii) a spatio-temporal closure for given activities as indirect reduction of effort and/or effort displacement on areas with lower availability. Different levels of compliance can be additionnally tested for each of these regimes. Laslty, the model acts as a support for functional response sub-models to test assumptions on fishermen response and effort re-allocation reacting to the implemented rules e.g. CPUE- versus cost for fishing- driven reallocation of effort in case of MPA application or regulated reduction of effort.

The population dynamic model for the both simulation tools was conditioned first on the Eastern Stock of the Baltic Cod having a SSB far below the reference points. Parametrization was done from the BITS survey to define a spatio-temporal age-disaggregated availability for the stock and/or from results from a hydrodynamical model for the Baltic Sea to define population zones and migration patterns and from the WGBFAS2007 to fit stock-recruitment, maturity-at-age and weight-at-age relationships. 'Good' and 'bad' forcing environments were identified and led to differenciate two set of values for population parameters depending on the years of inflow from the North Sea to the Baltic Sea.

The exploitation model including the initial allocation of effort was conditioned from the monthly- and ices square- disaggregated log-book data for the main countries participing in the Baltic Cod fishery (Denmark, Sweden, Latvia, Poland, Germany) and the missing effort from other unreported countries was completed to calibrate on the WGBFAS2007 eastern cod stock landings. Fleets were defined as a combination of country, vessel size and set of gears used, and specific spatio-temporal allocations of effort using a particular gear (i.e. metiers) were also explicitally modelled.

In the FLR version of the application, management regimes were defined to support the new F-adaptive approach preconised by the WGBFAS2007 for the recovery of the Baltic Cod defining a Harvest Control Rule based on a step by step reduction in overall F until the ultimate F_{target} is reached where the fishery is considered sustainable. The spatio-temporal closure effect from the 2006 commission proposal was also evaluated in comparison to a realistic total seasonal closure scenario. Presumable ways in re-allocation or displacement of the fishing effort in response to these regulations were investigated.

See also:

Except where otherwise noted, content on this wiki is licensed under the following license:CC Attribution-Noncommercial-Share Alike 3.0 Unported