We took as a starting point the hake – sole – Norway lobster Bay of Biscay ISIS-Fish database used for COSELMAR project16,20 (see http://isis-fish.org/download.html section “Bay of Biscay scenario dataset”, Database V0 in Fig. 1). This database was built using 2010 data, and was not calibrated, as it was designed for a geo-foresight study. Since our aim was to describe the system over a decade and simulate realistic dynamics close to available observations to assess management measures, we needed to update the parametrisation and calibrate the database. We took 2010–2012 as the calibration period, and 2013–2020 as the simulation period (grey arrow Fig. 1). The database has a monthly temporal resolution (constrained by the ISIS-Fish framework) and the spatial scale was set to match ICES statistical rectangles (0.5° latitude by 1° longitude rectangles, defined by the International Council for the Exploration of the Sea (ICES) https://www.ices.dk/data/maps/Pages/ICES-statistical-rectangles.aspx), consistent with available knowledge and data.
In this section, we firstly describe all the data sources used to update and calibrate the database. Then, for each main component of an ISIS-Fish database – i.e. populations, exploitation and management – we describe this paper’s database parameters and assumptions. We finally describe the calibration procedure (inspired by previous work21,22), of which some results are shown in the Technical Validation section. We summarized this workflow in Fig. 1.
Data sources
Data sources, estimates, and literature (including grey literature) were needed to update and calibrate the model. They are marked in Fig. 1 with salmon (data sources and estimates) and mustard (literature) blocks:
SACROIS23: French landings and effort logbook declarations for 2010 were made available at the log-event*commercial category*ICES statistical rectangle*population scale. It was used to design exploitation features of the database, as well as populations spatial structure.
LANGOLF survey: 2006–2010 LANGOLF surveys observations for 2006–2010 were made available for Norway lobster. They were used to work on Norway lobster abundance per length class and sex.
Intercatch: catch observations for 2010–2020 in the Bay of Biscay for hake, at the quarter-métier group scale, and catch observations per class for sole on 2010–2012, and 2010 Norway lobster catch observations per sex and length class24, used to describe the inter-annual effort dynamics, to calibrate and validate the model.
Estimates of hake abundance per size class in 2010, and hake quarterly estimates of recruitment on 2010–2012 from a northern hake spatial stock assessment model21, used to inform hake biology assumptions (named Other 1 in Fig. 1).
ICES WGBIE24 2010 estimates of abundance per class (sole and Norway lobster), to inform their abundance at the initial time step; 2010–2012 yearly fishing mortality estimates per age class (sole) to calibrate the database (named Other 2 in Fig. 1).
Other population, exploitation and management assumptions were informed with scientific literature25 and grey literature26,27 (Literature block in Fig. 1).
Management assumptions were informed with legal texts2,4,28,29,30,31,32,33,34 and reported quota values in working group reports24.
About populations
This section describes for each species the assumptions and parameters values, except for accessibility, which has been calibrated, as described in section Calibration procedure. For all assumptions and values, more details are provided in Supplementary Information’s section 2.2.
Hake
The stock size structure was defined with 1 cm size bins for [1;40[cm individuals, 2 cm for [40;100[cm individuals, and 10 cm for [100;130+] cm individuals35. Areas of presence were defined based on 2010 SACROIS French landings data per commercial category and statistical rectangle23, leading to the definition of a presence, a recruitment, an interim recruitment and a spawning area25 (see Supplementary Information’s section 2.2 and Figure S1). These areas allow for the description of intra-Bay of Biscay migrations related to spawning and recruitment processes: mature individuals aggregate at the beginning of the year on the shelf break to spawn, and then disperse on the shelf36,37,38,39,40 (at the beginning of April and July in the model). Also, from age 1 (around 20 cm), individuals in recruitment zone spread in interim recruitment zone, to model a diffusion towards areas neighbouring the nursery area, at the beginning of each time step (see Supplementary Information’s section 2.2 and Table S11). Maturity-at-size and weight-at-length relationships were the same functions as used by ICES working group35,41. Natural mortality was fixed at 0.5, basing on preliminar runs, instead of the commonly used 0.442. Recruitment values were defined prior to the simulation for 2010–2020 using available estimates on the 2010–2015 time series21,27. Deterministic estimates from these sources were allocated to the recruitment area in the Bay of Biscay and the beginning of each month in January-September on the whole time series, of which values are provided in the Supplementary Information’s section 2.2 and Table S3. Growth is modelled through monthly growth increments5,25. However, given the different widths of size bins in the implemented size structure, a correction was provided to values in the transition matrix to eliminate artifacts when growing to a size bin wider than the size bin of origin, as detailed in Supplementary Information’s section 2.2. Abundance at the initial step in each zone was estimated from Bay of Biscay abundance estimates for 201021. Mature individuals over 20 cm were allocated to the spawning area, all individuals strictly shorter than 20 cm were allocated to the recruitment area (as they were assumed to be less than 1 year old), and remaining individuals were allocated to the interim recruitment area. None were allocated to the presence area, in which individuals will go later in the time series, after disaggregating from the spawning area25 (Table S13).
Sole
The stock is age structured, with 7 classes going from ages 2 to 7+43 (Table S2). No seasonal variations were implemented. Only a single presence zone was defined (see Supplementary Information’s section 2.2 and Figure S1), as in preliminary runs defining more presence areas for sole did not yield more knowledge in this study. We implemented ICES working group values for natural mortality, weight-at-age (Table S1) and maturity-at-age43. Recruitment occurs at the beginning of each year, individuals being recruited at age 2 (ages 0–1 were not modelled; Table S4). We implemented ICES working group estimates27 for abundance at initial time step (Table S14).
Norway lobster
The stock has a sex-size structure, with 1 mixed recruitment class at 0 cm; 33 length classes for males at 2 carapace length mm intervals, from [10;12[to [72;74[carapace length mm; 23 length classes for females at 2 carapace length mm intervals, from [10;12[to [52;54[carapace length mm. A single presence area was defined: the Great Mudbank21 (see Supplementary Information’s section 2.2 and Figure S1). Several seasonal processes occur for this stock, impacting recruitment, accessibility and growth: 1/ January, begins with the annual recruitment. Females are inside their burrows, less accessible; 2/ February-March females are inside their burrows, less accessible; 3/ April: Spring moulting, females are more accessible; 4–5/ May-August females are more accessible; 6/ September, females are inside their burrows, less accessible; 7/ October: Autumn moulting only for immature females and all males, females are inside their burrows, less accessible; 8/ November-December, females are inside their burrows, less accessible44. We implemented ICES working group values for natural mortality, weight-at-class and maturity-at-class45,46,47. Growth occurs twice a year, when moulting in April and October, and is modelled with growth increments. Recruitment occurs at the beginning of each year, modelled with a Beverton-Holt relationship26, and was assumed to have the same spatial distribution as spawning stock biomass. Abundance at initial step was derived from LANGOLF survey observations and ICES WGBIE estimates25,26 (Table S16).
About exploitation
The fishing exploitation structure (fleets, strategies, métiers and gears) were derived following a classification method on SACROIS 2010 landings and effort data13,23 from French fleets, and taken from a TECTAC project (https://cordis.europa.eu/project/id/Q5RS-2002-01291) database for Spanish trawlers. More details on their definition are provided in Supplementary Information’s section 2.3, Tables S5–S9 and S20–S21 and Figure S3. Spanish longliners and gillnetters fleets exploitation was described based on catch (observations from Intercatch48) rather than effort.
Hake selectivity and discarding functions (one for each gear) were taken from estimates of a spatial hake stock assessment model21. Parameters values and formulæ are provided in Supplementary Information’s section 2.3 and Tables S6-S7. On top of this, inter-annual fleet dynamics factors were included in equation (21) of ISIS-Fish documentation8 in order to account for observed catch temporal variations. These factors are therefore multiplicative parameters of the target factor of each species for each métier. They are computed using observed catch27 and differ according to the period and targeted species:
over 2010–2016, it is a ratio of observed catch in weight per year over catch observations for 2010: for hake, one per métier *season*year (left(frac{ObservedCatc{h}_{metier,season,year}}{ObservedCatc{h}_{metier,season,2010}}right)), for sole, one per métier *year (left(frac{ObservedCatc{h}_{metier,year}}{ObservedCatc{h}_{metier,2010}}right)), and for Norway lobster, one per year (identical for each métier catching Norway lobster) (left(frac{ObservedCatc{h}_{year}}{ObservedCatc{h}_{2010}}right));
over 2017–2020: at the time of writing these assumptions, more recent data was not available, and ratios were deduced from trends on 2014–2016. A linear model was fitted on ratios deduced earlier on 2014–2016. If a significant trend was identified (hake: whitefish trawlers quarters 2 and 4, longliners and gillnetters seasons 2–3; sole and Norway lobster: all métiers), the slope was used to deduce 2017–2020 ratios (the slope was halved for hake whitefish trawlers and sole and Norway lobster values to avoid unrealistic high values of effort). Otherwise, 2016 ratios were used.
All values are provided in Supplementary Information’s section A.2 Tables S22–S24, and the final values of target factors are derived from the Calibration procedure.
About management
We implemented a set of management rules close to what is currently implemented in the Bay of Biscay.
All stocks are managed by TALs (Total Allowable Landings) until 2015 and then by TACs (Total Allowable Catch), except for Norway lobster, managed by TALs on the whole time series, not being under the landings obligation. To favour a better parametrisation, allowing for more reliable dynamics on the following years of the time series, no TALs were implemented during the calibration period (2010–2012; Fig. 1). These regulations were implemented from 2013 using historically TALs and TACs values24.
Landings of the three stocks are also constrained by a Minimum Conservation Reference Size regulation that was implemented for all stocks using values currently enforced in the studied fishery28. Likewise, from 2016, the Landings Obligation was implemented, with de minimis exemptions for hake and sole, depending on the year and the gear used to fish them2,31,32,33,34. See Supplementary Information’s sections 2.4 and A.3, Figure S2 and Table S10 for further details on these restrictions.
In response to the above management rules, a fishers’ behaviour algorithm has been developed to describe fishermen adaptation. Some métiers may be forbidden, depending on some conditions – the catch quota has been reached, the landings obligation is enforced – but also some values – the proportion of discarded catch, and also catch on previous years. Therefore fishermen change métiers within their strategy métiers set through a re-allocation of fishing effort to the latter set. This re-allocation aims to avoid quota overshooting. Further details about this algorithm are provided in the Supplementary Information’s sections 2.4 and A.3 and Figure S2.
Calibration procedure
The model has been calibrated using two parameters (population accessibility and fishing target factor) involved in the catchability process (equation (21) in ISIS-Fish documentation8). The objective of the calibration is to reproduce the dynamics of catch over 2010–2012 at the species*métiers group scale, for each year or quarter depending on available data’s granularity. Calibration is sequentially performed: accessibility parameters for each population were estimated first followed by the target factors. The estimation of each parameter set (parameter type * population) combination was separated, and values were estimated jointly within each parameter set. To account for the specificity of each population model dynamics (global age-based for sole, spatial and size-based for hake, spatial, sex and size-based for Norway lobster), an objective function is defined for each population to calibrate their accessibility. More details on objective functions and procedures are provided in Supplementary Information’s section 2.5, as well as estimated values in Tables S17–S19.
Hake accessibility
The calibration for hake accessibility is based on a procedure developed for a former version of the database25. One parameter was estimated per quarter, all values being equal across length classes. The model outputs were fitted to hake catch observations in weight in the Bay of Biscay in 2010–2012 per length class.
Sole accessibility
One parameter was estimated per age class. The model outputs were fitted to WGBIE fishing mortality per age class for sole27 in 2010–2012.
Norway lobster accessibility
One parameter was calibrated per sex and length class. The model outputs were fitted to catch in numbers per length class and sex in 2010 per quarter provided by WGBIE.
About target factors
Target factors drive how the effort is distributed between populations, métiers and season*year combinations. They were split in 3 components: a fixed component derived from the SACROIS effort dataset analysis (Tables S25–S27), another fixed component driving inter-annual variations of fishing effort (Tables S22–S24), derived from catch observations, and finally an estimated component (Tables S28–S30), allowing to tune the model’s dynamics to observed catch. This section focuses on the estimation of the latter.
Hake target factors
20 parameters were defined, for each combination of the 5 groups of métiers (longliners, gillnetters, whitefish trawler (coastal), whitefish trawler (not coastal), Norway lobster trawler, see definition Table S8) and 4 quarters. We fitted the model’s outputs to the same data and with the same objective function as for hake’s accessibilities estimation.
Sole target factors
1 estimated component per group of métiers (gillnetters, Norway lobster trawlers and whitefish trawlers) and quarter. We fitted the model’s outputs to sole catch in weight on 2010–2012 for each métier and quarter.
Norway lobster target factors
1 estimated component per group of métiers (Norway lobster trawlers and whitefish trawlers). We fitted the model’s outputs to monthly Norway lobster landings data per length and sex class for 2010.
Base simulation
The base simulation ran from January 2010 to December 2020 inclusive, with a monthly time step, using the database and parameters values described in this document. Several outputs of interest may be explored after a run: catch (discards and landings), as done in several figures in this paper, but also biomass (total biomass or mature biomass), fishing mortality values, or effort, all at a fine spatio-temporal scale.
Source: Ecology - nature.com
