Quick link to: EFIMAS home, WP4, CS6

FLR approach for the Case study 6

See General introduction to the CS


Description of the fishery, stocks and management system

See General introduction to the CS

Description of the base case and scenario evaluations

See General introduction to the CS

Data and parameters

Stock data

Catch-at-age and biological data are taken from The Working Group of Hake, Monkfish and Megrim (WGHMM). Indices of abundance are simulated based on the XSA output

Parameter Source of information Format
non spatial biological parameters : M, weights growth Assessment WG Literature Report articles
reproduction, recruitment parameters : zoning, timing, fecundity, maturity Literature Survey data articles data files
spatial parameters spatial distribution per age/length group catchability literature Assessment WG articles report

Fleet data

Data on catch composition by stock and fleet are taken from both WGHMM and AZTI-Tecnalia database.Economic data is obtained from SEAS (Statistical Service of The Department of Agriculture, Fisheries and Food of The Basque Government)

Parameter Source of information Format
vessels characteristics TECTAC results and data report data files
Metiers Analysis of logbook and activity data, with auxiliary information from survey data data files
fleet dynamics TECTAC results and data existing interviews and studies reports data files

Economic data (incomes, costs and prices)

Data of all economic variables are available. Consistent with the DCR Variables to be used as key factors are: Variable costs including fuel costs. Indicators to be used as key factors are: Gross margin, Gross value added, Gross cash flow, Net profit.

Parameter Source of information Format
Prices (species)By size By country By gear Fuel prices Official statistics Eurostat (hake imports and quantities by origin)Merca Barcelona MERCASA ICEX Auctions reports data files
Turnover TECTAC results and data ESESPV results and data (survey)Sales notes and landing declarations report data files
Subsidies (production) Official reports (BOE, etc.) reports
Labour costs TECTAC results and data reports
Wages & salaries Social security ESESPV results and data (survey)Official statistics surveys Sales notes and landing declarations data files
Fuel cost Accounts Surveys/questionnaires TECTAC results and data ESESPV results and data (survey) reports data files
Other running costs Variable :Landing costs Sales Lubricant Ice Non-variable : Maintenance and repair (vessel and gear) Insurance Accounts Surveys/questionnaires TECTAC results and data ESESPV results and data (survey) reports data files
Fixed costs (capital costs): Depreciation Literature (try to define precise and comparable indicators)Accounts Surveys/questionnaires articles reports data files
Investment Capital value Fishing rights Literature (try to define precise and comparable indicators) Accounts Surveys/questionnaires articles reports data files


FLR (Operating Model)

The operating models should ideally be parameterized using raw data but in many fisheries systems the only available data is that used in the assessment. The ‘caaOM()’ is an R function, built using FLR libraries, that aims to describe a fishery system and can be easily parameterize using the output and input of an assessment model such as XSA. But if other data sources are available they can be used as well. The functions consist in a biological and fishery operating model, a observation model to feed the assessment model, FLXSA, a management procedure based in a Harvest Control Rule which returns an annual or a multiannual TAC and an implementation model.

Given an initial vector of numbers at age and some biological and technical parameters the population is carried forward using the usual survival and catch equations, effort-fishing-catchability relationship and the specified stock-recruitment model.

The operating model, the part of the function that simulates the ‘true’ population and the fishery, is split in two parts, the historic part and the projection part. In the historic part of the model the effort of the fleets can be given as input data for each fleet or estimated within the simulation conditioned to produce the observed catch-at-age matrix. In the following year to last year of historic data, an assessment is carried out and the specified harvest control rule (HCR) is applied to calculate a TAC for the following years. The TAC is shared among fleets, the proportion of the TAC belonging to each fleet is considered constant over the years and must be given as input data. This process of assessment and HCR application is replicated in each projection year. The efforts of the fleets in the projection are estimated conditioning on the proposed TAC.

Packages used are FLCore Ver 2.2.1, FLAssess FLXSA and FLSTF

Data inputs are modelled by the FLCore classes FLStock, FLIndex, FLIndices, FLBiol, FLFleet and FLFleets, FLSR.

which contain almost all the input data required for and the outputs from the conditioning.

Conditioning of Operating Model

The Northern Hake base case has been defined following the assumptions of the assessment working group so in this case the OM is parameterized using only assessment data. So the biological parameters used in both ‘real’ and ‘observed’ are the same. The different stock-recruitment relationships considered have been parameterized fitting the different models to the SSB and recruitment data estimated by the WGHMM. The variance of the error around stock-recruitment curve is also taken from the fitting to the model.

The unique fleet considered in the Base Case has been parameterized in two different ways, in the first one the historical fishing mortality of the fleet is set equal to the one estimated in the WGHMM using XSA, and in the second one using a catchability proportional to the former fishing mortality the effort in each year is estimated conditioned to produce the observed catch.

In an alternative case a population which grows twice faster than the base case one is considered, the parameters for this growth pattern are also taken from the WGHMM report in which a simulation study with this different grow pattern is carried out since 2004. Different values for natural morality are going to be tested in alternative cases but there is no scientific basis for the choice of a specific value.

The fleet specific catchabilities in the alternative case has been estimated using annual effort estimates and assuming that the total fishing mortality is divided among fleets in the same proportion as the catch. Then the catchability is calculated using the effort-fishing mortality-catchability relationship.

Economic data for conditioning the economic data of the fleets is obtained from the statistic service of the department of agriculture, fisheries and food of the Basque Government. It includes data from the vessel and the firm it self and is composed of all the variable, fixed, financial and depreciation costs.

Historical Estimates of Time Series

To simulate the population a single species operating model have been developed. It uses the survival and catch equations to project a given initial population forward. Effort in each year is conditioned to produce the input catch at age in the historic part of the simulation and the settled TAC in the projection part.

hke.1 N. Hake Base case WG XSA, Recovery plan HCR Segmented regression SR a
hke.2 N. Hake Base case WG XSA, Recovery plan HCR. Segmented regression Blim SR a
hke.3 N. Hake Base case WG XSA, Recovery plan HCR. Ricker SR a
hke.4 N. Hake Base case WG XSA, Recovery plan HCR. Beverton and Holt SR a
hke.5 N. Hake Base case WG XSA, Recovery plan HCR Segmented regression SR. Conditioned to CAA. a
hke.6 N. Hake Base case WG XSA, Recovery plan HCR. Segmented regression Blim SR. Conditioned to CAA. a
hke.7 N. Hake Base case WG XSA, Recovery plan HCR. Ricker SR. Conditioned to CAA. a
hke.8 N. Hake Base case WG XSA, Recovery plan HCR. Beverton and Holt SR. Conditioned to CAA. a
hke.9 N. Hake Base case WG XSA, Recovery plan HCR Segmented regression SR.Using WGHMM F estimates. a
hke.10 N. Hake Base case WG XSA, Recovery plan HCR. Segmented regression Blim SR. Using WGHMM F estimates. a
hke.11 N. Hake Base case WG XSA, Recovery plan HCR. Ricker SR. Using WGHMM F estimates. a
hke.13 N. Hake Base case WG XSA, Recovery plan HCR. Beverton and Holt SR. Using WGHMM F estimates. a

Plots of SSB, Recruitment, catch and Fishing mortality were made for each run.

Biological Parameters: Recruitment, Growth

The recruitment is simulated using a stock-recruitment relationship, all the models implemented in FLSR class can be used, at the moment this class includes 4 different SR models, Beverton and Holt, Ricker, Segmented Regression and Quadratic hockey Stick, but it allows to define new models.

Double growth of hake has been tested for two different SRR. (not working yet).

average Segmented regression Blim Single
high Segmented regression Blim Double
average Ricker Single
high Ricker Double

Uncertainty can be introduced in all the biological parameters, just filling the input FLBiol object with different parameters along the iterations. The same can be done with the SR model parameters in the FLSR object but additionally there are another two ways for introducing uncertainty in the SR process. The first one is to fill the ‘var’ slot of the object, in which case a multiplicative error is added to the recruitment estimate in each of the iterations, the error is sampled from a lognormal distribution. The second option is to add an error sampled from the residuals stored in the ‘residuals’ slot of the object.

Fisheries & Fleets

The link between biological and the fishery operating models is done via the fishing mortality. As usual the fishing mortality is assumed to be equal to age specific catchability times effort. The catchability is an input parameter and as with the biological parameters it can vary along iterations just filling the catchability slot with different value per iteration.

Fleet dynamic is include in a Discret model approach. In the base case there is only one fleet exploiting the resource so neither fishery dynamics nor economics are considered. In alternative cases the fishery is divided in several fleets. The dynamics of the fleet is simulated analyzing partially the short term behavior of them. Within a fleet the choice between métiers is modelated using a random utility model (RUM) which parameters are calculated externally. Some of the observables of the RUM, such as the TAC of other species, are given as input data and other are calculated using the simulated data.


The economic variables to be used as key factors are, mainly the variable costs including fuel costs. In order to evaluate the economic performance of the fleet several short /long term indicators are to be calculated. • Gross margin, • Gross value added, • Gross cash flow, • Net profit.

Price evolution has been also tested. Given that supply is very inelastic in the short run and the producers are virtually price takers, the inverse demand appears to be a very natural model for the price formation. Thus, inverse demand functions have been estimated for fresh hake caught with two different gears, long-line and trawler. The analysis used 56 monthly ex-vessel price and quantity observations of Galician long-line and trawler fresh hake market, for the period January 2001 until August 2005. To check if the demand model is price or quantity dependent a Pairwise Granger Causality Test has been performed. This test did not provide a clear evidence of which variable, price or quantity, was exogenous and endogenous. So, following the most common criteria that quantities better explain prices for seafood products, the quantity was considered the exogenous variable. Flexibilities for the Galician long-line hake heve been estimated using wide range of linear and double log single equation models where prices and total expenses are incorporated in both real (deflated) and nominal (non-deflated) values, while for the trawler hake only a double log single equation demand models have been estimated. As seasonal and autocorrelation effects were found to be significant they were incorporated into the models. The estimated equations are: (1) Galician ex-vessel market for long-line fresh hake (Std. Errors in parenthesis)

LNPLL = -0.141817(0.013967)-0.877992*LNQLL (0.116562)+1.077147*LNGASTOGAL(0.088486)-0.191541*LNQTR(0.034249)-0.063637*LNQTRL (0.395422)+0.605471*AR(1)(0.041252)++0.239991*R(12)(0.032981)-0.926496*MA(12)(0.017808)

R-squared: 0.977746

Adjusted R-squared: 0.973419

F-statistic: 225.9604 Prob(F-statistic): 0.000000

(2) Galician ex-vessel market for trawler fresh hake (Std. Errors in parenthesis)

LNPTR = -2.352181(1.0177500)+0.015215*LNQTR (0.045121)+0.464410*LNGASTOGAL(0.129491) -0.183760*LNQLL(0.085139) -0.067777*LNQTRL(0.019984)-0.183132*AR(1) (0.151938)-0.136682*AR(12)(0.124737) +0.933030*MA(12)(0.028210)

R-squared: 0.675758

Adjusted R-squared: 0.612711

F-statistic: 10.71830 Prob(F-statistic): 0.000000

Where: LNPLL: Price of fresh hake caught by long-liners in the Galician ex-vessel market. (in logs) LNQLL: Quantity of fresh hake caught by long-liners in the Galician ex-vessel market. (in logs) LNGASTOGAL: Total expenses in all hake products commercialised in the Galician ex-vessels markets. (in logs) LNQTR: Quantity of fresh hake caught by trawlers in the Galician ex-vessel market. (in logs) LNQTRL: Quantity of fresh hake caught by littoral trawlers in the Galician ex-vessel market. (in logs) AR(1): Autocorrelation of order 1 AR(12): Autocorrelation of order 12 MA(12): Moving average of order 12

Results show that the own price and the scale flexibilities are sometimes not significant. So, further analysis should be performed, as well as it would be interesting to obtain longer or more relevant series for the analysis. It implies that for this first attempt, price will not respond to a model.

See complete information in relation to price flexibility for Northern Hake

Reference Points

Reference points are those used in the HCR. Blim=100.000 tonnes Bpa=140.000 tonnes Flim=0.25 BBFPA=0.35

Net profit >0

Observation Error Model

The observables in the model are the abundance indices and the catch-at-age matrix. The abundance indices can be generated using a ‘q-Model’ or a ‘power-Model’ with the given catchabilities and parameters for each index, other catchability models can be easily added to the algorithm. The error in the indices can be introduced externally, specifying a different catchability in each of the iterations, or adding a multiplicative lognormal error to each index. The error in the catch-at-age matrix is introduced using a multinomial distribution, the parameters of this distribution must be given as input data and represent the proportion of individuals of one age group that goes into another group by mistake.

The error around stock-recruitment curve can be modeled using either bootstrap or Monte Carlo methods. The bootstrapping is done from the residuals stored in the FLSR object. To avoid high recruitments when the spawning stock biomass is low, residuals can be divided in two groups depending if the corresponding biomass is greater or smaller than a given threshold biomass. Then, when predicting the recruitment, the bootstrap is done from one group if the current ssb is greater than the threshold level and from the other one otherwise. Other option is to model uncertainty using a lognormal error around the stock-recruitment curve, in this case, a number is drawn randomly from a lognormal distribution with zero mean and the specified standard error.

The error in the catch-at-age matrix is simulated using a multinomial distribution. When calling the OM a square matrix must be supplied with one column per age, each column representing the mean proportion of individuals of one age, that goes into the age corresponding to the column by mistake (aging errors, sampling errors…). The difficulty here lies in the quantification of this errors. The uncertainty in the natural mortality can be introduced using lognormal or uniform distributions. The distribution shape must be the same for all ages but its parameters can be age specific. In each iteration a random vector is drawn from the chosen distribution and is considered constant along years,so no interannual variability in natural mortality is considered. An observation error can be introduced in the simulated abundance indices drawing a random number from lognormal distribution with zero mean and specified standard error. The standard error can be age dependent.

Implementation error

Implementation error is modeled in a very simple way simulating only the proportion in which the TAC is exceeded. The input vector, λ = (λ1, λ2, …,λn), represents the percent in which the settled TAC is exceed or not reached in years i=1,..n. The algorithm also allows using a matrix with one λ vector per iteration, so different distributions for λ can be tested

Management Procedure

A) The Recovery Plan

The management procedure in the base case is the actual recovery plan of Northern Hake, which established the TAC according to the following Harvest Control Rule,

• 2004: The TAC will be that corresponding to a fishing mortality equal to Fpa= 0.25.

• 2005 onwards:

o The F corresponding to the TAC should not exceed Fpa.

o The SSB predicted for next year using the TAC should not be smaller than the last estimated SSB.

If following the previous two rules, the resulting TAC is 15% greater or smaller than the previous TAC, the TAC should be set to be equal to +/- 15% of the previous TAC. If the predicted SSB using the TAC obtained applying the last rule leads to a SSB lower than Blim = 100000 the TAC should be decrease to a level that ensures that the future SSB is greater than Blim.

Multi-annual TAC:

B) Multiannual TACs

The use of multi-annual TAC is something that is still under revision by many fisheries managers. Greater stability for fishermen is always welcome and it seems that the “race to fish” should be reduced by fitting to the extraction limits adopted. The idea of this system is to be able to predict now the future TACs. Defining a TAC for several years has the problem that future evolution of the stocks has to be predicted in the present. Simply, this is the main risk.

Some diferent scenarios has been tested:

hke.14 N. Hake Alternative case hke.2, Multi-annual (4,5,6,7,8, years) HCR a
hke.15 N. Hake Alternative case hke.2, Multi-annual (4 years), F costant HCR a
hke.16 N. Hake Alternative case hke.2, Multi-annual (4,5,6,7,8, years) HCR, no bounds a

Apart from the idea of giving a deterministic future to the industry, there should be other reasons for setting this management procedure. Compliance is a key factor in the introduction of this multi-annual TAC. Levels of quota violations in fisheries appeared to be driven mainly by financial incentives. Multi-annual TAC reduces these financial constraints, giving the industry a stable perspective for the future. The coded function allows defining multiannual HCRs that returns a set of TACs for several years, which will permit to analyze the interaction between multiannual TACs and compliance.


Background to understand the importance of the ITQ model for Spain

Spain was one of the mayors fishing power in the world, even if nowadays its importance has decrease. The Spanish fishing sector is the biggest of the EU-15 in terms of landings value and direct employment and the second in terms of landings in volume (after Denmark) and fish per capita consumption (after Portugal). All these conditionings have their mirror in the total fishing power and on the regional distribution of this fishing power where after Galicia, the second mayor Spanish fishing region is, in terms of total GT, the Basque Country.

The Basque trawl fleet (the one that has been analyzed in this paper in terms of management), belongs to what is known as the “300 fleet”. This fleet is composed of bottom otter trawlers (“Baka”and “Bou”), and bottom Very High Vertical Opening (VHVO) pair and trio trawlers, which operate in ICES subareas VI, VII and divisions VIIIabd. Nowadays in the Basque country only “Baka” otter trawlers and VHVO pair bottom trawlers can be found.

Spain entered the former EEC in 1986 assuming all the European regulation refereeing the fishing sector, in which they can be found output (TAC,…) and input (TAE,…) regulations.

The evolution of the management measures, have clearly affect the shape of the total and specially the Basque part of the 300 fleet. It has change, obviously, in terms of the number of vessels (from more than 90 vessels to less than 50) but also in the characteristics, and in the fisheries targeted by these vessels. Certainly, all these changes are not only derived from the evolution and changes in the regulation, the evolution of the target stocks and the technological change has also affected the evolution of the fleet.

The main technical measure implemented to this fleet has been the limitation in the minimum mesh size (to be precise, the regulation of the mesh opening) that can be used. In that there always have been two different areas. One covered by Sub-area VIII (except VIIIc) which has had a less restrictive mesh size limitation that Sub-areas VII and VI.

The TAC was first implemented when Spain joined the EU in 1986. Setting TACs involves the fixing of maximum quantities of fish that can be caught from a specific stock over a given period of time. This operation requires cooperation among the various parties enabling those involved to come to an agreement regarding TACs and an allocation key for sharing them. The EU went on to share fishing opportunities in the form of quotas among Member States. A formula was devised to divide TACs according to a number of factors, including countries' past catch record. This formula is still used today, on the basis of what is known as the principle of 'relative stability' which ensures Member States a fixed percentage share of fishing opportunities for commercial species. Even if the share has been maintained stable over time, the growing scarcity of the key stocks has eroded significantly the fishing opportunities for these fleets.

The TAE is previous to the TAC regulation. The Spanish offshore fleet in 1977 (when the exclusive economic zone was increased to 200 miles) was composed of something between 500 and 600 vessels. The fishery was regulated by the mean of licences where each vessel had a single licence (which allowed, “theoretically”, to around one month fishing). This situation changed in 1981, when it was decided to create a list with all vessels operating in these fisheries, in order to create the access rights to these fisheries (a single fishing right per vessel). Hence, the number of vessel with fishing rights was 416, and it diminished to 300 when Spain entered the EU, while the total fishing rights were still 416.

The idea was not to maintain the number of vessel stable, but to maintain the fishing rights even if the number of vessel decreased. The system changed dramatically when Spain joined the EU. In this period the number of vessels in that list was close to 300 and the so-called “300 list” was created, but only 145 could operate simultaneously. And they changed from fishing rights by area to access coefficients using the average power of the fleet in that period (552Kw or 750 hp)

The problem with this system was that if a vessel with more power sold its fishing days to a less powerful one, the access coefficient was not exhausted, and the surplus could be shared among the remaining vessel of the company. In other words the more powerful the vessel, the more costly were the fishing rights.

There was also another important change, before 1997 these rights were always attached to the vessel and there were only accumulated to other vessel when it disappears from the list (decommissioning, sunk,). But in 1997 the scope change and the possibilities of transference of fishing rights increased. Hence the possibility of reallocate within the firm or to transfer to some other firms was opened with two limits. Each vessel had to have at least 200 days at sea and no more than 315 days taking into account all the areas. This system continued until 2003. Currently there are several negotiations going on to decide whether this system will continue or not.

Data requirements and base knowledge for modelling Individual Transferable Quotas.

This report aims to cover a first step on the modelling and data requirements in order to discuss the possibilities of effective implementation of Individual Transferable Quota (ITQ) management strategy in the EFIMAS Northern hake case study. Beginning with some clarifications related to the nature of the property, the TAC setting criteria and the need to deal with the species inclusion-exclusion, some issues related to eligibility, initial allocation and transferability rules are pointed out. Special attention is paid to showing a potential procedure to fill in the quota market box in the simulation framework based of econometric estimation of each vessel’s derived demand of quota in a dual framework. Once the data and the modelling requirements have been analysised, the main conclusion of the report is that when aiming to model fleet behaviour under an ITQ regime the operating model should be based on individual vessels instead of fleets. Therefore, unfortunately the agregation level of the northern hake model developed under FLR approach does not allow to properly analyse an alternative management scenario based on ITQs.

See complete text in UPV ITQs.pdf

See complete text in UPV document on Individual Transferable Quotas, Base Knowledge for modeling the Northern Hake Case Study


Brief view of OM development

The operating model is already developed and tested for the Northern Hake case study of EFIMAS and COMMIT, using FLR libraries. The model has been coded in a generic way and in principle any species assessed using XSA or similar age-structured methods can be simulated using it. It describes a biological population exploited by a set of fleets and managed applying an annual harvest control rule. The assessment is presently carried out using eXtended Survivors Analysis using the FLXSA library

A brief scheme of the OM and example of the simulation of the CS6 is presented in A generic OM development

Stock Recruitment Relationship results of dimulations

Beverton and Holt has been rejected as a suitable SRR. After the analysis carried out the best fit, acording to AIC criterion, is Ricker. On the other hand, it gives the less “optimistic” (lowest recruitment levels) scenario.

For the rest of the scenarios, segmented regression's AIC is similar to Ricker. It performs the best fit with what observed in WGHMM. Segmented regression with Blim, has a clear biologic significance and is the most “optmistic” highest recruitment levels scenario

Management strategies

N. Hake Revovery plan scenario:

It appears to works only in the short run. It performs well as a “recovery tool”, but in the long run the TAC reachs such a high level that it has to be reduced. In this Long Term Run the inestiability of TAC is remarkable.

Multi-Annual TAC scenarios:

In the preliminary trials it appears that results on a stable TAC and a stable biomass.

This Link points strait to COMMIT WiKi Page. LOL

efimas1/wp4/cs6/appr1/main.txt · Last modified: 2008/06/13 07:53 by ms
Except where otherwise noted, content on this wiki is licensed under the following license:CC Attribution-Noncommercial-Share Alike 3.0 Unported
Recent changes RSS feed Donate Powered by PHP Valid XHTML 1.0 Valid CSS Driven by DokuWiki