General description of the tool

According to Article 15 of the Common Fisheries Policy (Regulation EU No 1380/2013), the Landing Obligation in the Mediterranean applies to catches of species which are subject to minimum sizes as defined in Annex III of Regulation (EC) No 1967/2006 (Table 1). The Mediterranean demersal fisheries are both multifleet and multispecific, with more than 100 species in their landings (Moranta et al. 2008). In addition, both small-scale and bottom trawl fisheries are characterized by spatial and temporal variability of their fishing strategies that mainly depend on the bathymetric range and determine both the target species and demersal communities exploited (Colloca et al. 2003, Massuti and Renones 2005).

In the Balearic Islands (western Mediterranean), commercial trawlers use up to four different fishing tactics (Palmer et al. 2009), which are associated with the shallow and deep continental shelf, and the upper and middle continental slope (Guijarro and Massuti 2006, Ordines et al. 2006). Vessels mainly target striped red mullet (Mullus surmuletus) and European hake (Merluccius merluccius) on the shallow and deep shelf respectively. However, these two target species are caught along with a large variety of fish and cephalopod species. The Norway lobster (Nephrops norvegicus) and the red shrimp (Aristeus antennatus) are the main target species on the upper and middle slope respectively. The Norway lobster is caught at the same time as a large number of other fish and crustacean species, but the red shrimp fishery is the only Mediterranean bottom trawl fishery that could be considered monospecific.

Stakeholders from the Balearic Islands consider that the best measures to reduce discards in the Mediterranean fisheries are the improvement of gear selectivity and the use of spatiotemporal closures for effort control (Task 2.5). They also consider that in order to protect recruitment, both the fishermen skill and the available scientific knowledge can allow identifying the best seasons, depths and/or areas to be closed for some target species of the bottom trawl fishery. Such considerations are in line with the main aim of this task, which is to provide decision support tools to assist fishers to make choices of fishing location to avoid discards. This work develops the scientific information on fish distributions in time and space, nursery areas and discarding hotspots from the Balearic Islands (Western Mediterranean Case Study) bottom trawl fishery.

Input data

Species analysed

A total of 19 demersal species, including fish, crustaceans and cephalopods, were chosen for the analyses (Table 2). Together with the species regulated by MLS in the Mediterranean, the most important commercial species for the bottom trawl fishery from the Balearic Islands were studied. Analyses to delineate hotspots for these species were done considering two different criteria for the definition of recruits and adults: 1) MLS in the case of the species under regulation; and 2) size at first maturity (L50) for those species without MLS but also for those having MLS.

Data sources and analyses

Two different data sources were used: 1) Fishery independent data taken during scientific surveys (MEDITS); and 2) Fishery dependent data collected by observers on board bottom trawlers working under commercial conditions. These two data sources were analysed separately due to differences in the spatiotemporal sampling. Whereas surveys only provide information of the specific period when they are performed, the monitoring of the fishery along the year allows analysing seasonal changes. Whenever possible (see Results), all study species were analyzed from samples taken by these two different sampling sources.

Fishery independent data: MEDITS surveys

A two step methodological approach following (Colloca et al. 2015) were used in order to analyze the spatial distribution of species density (number of individuals per Km 2 ). The methodology started with the identification of areas with higher densities (hotspots) on an annual basis, and then analyzed the persistence of these hotspots through the time series as indicative of the most important areas of density that are independent of interannual variability. Finally, the overlap of persistent hotspots for the different fish species was also analyzed.

Data was collected during the MEDITS scientific trawl surveys, which are annually conducted in spring-summer in European Mediterranean waters (Bertrand et al. 2002). The study area covered the continental shelf and upper slope fishing grounds between 50 and 800 m depth around the Balearic Islands (Western Mediterranean, Fig. 1). A total of 766 bottom trawls conducted between 2003 and 2016 were analyzed. Although the surveys are carried out from 2001, only data collected between 2003 and 2016 were used in order to analyse the same time period for the two data sources (information from onboard samplings is only available from 2003 on). Fig. 1. Location of the sampling stations analysed annually during the MEDITS surveys carried out around the Balearic Islands.

On a first step, the spatial distribution of the study species was analyzed considering, as aforementioned, different population fractions (under and above the MLS and the L50). The spatial modelling was performed with generalized additive models (GAMs; (Wood 2006) using the “mgcv” R-package. The smoothing parameters were selected by restricted maximum likelihood (REML), which is supposed to be more effective than other available options (Marra and Wood 2011). Depth, position (longitude, latitude) and year were used as explanatory variables in the GAMs.

Owing to the high frequency of zeros in the data, a two-stage GAM approach was used (Barry and Welsh 2002, Borchers et al. 1997). This approach assumes independence between presence- absence and abundance values, so that the likelihoods are obtained by multiplying the predictions of the two components (presence-absence and abundance). In the first stage, a presence-absence variable is constructed by giving a value of 1 to observations where the response variable is greater than zero and a value of 0 otherwise. The derived variable is modelled as a binomial distribution with a logit link function. In the second stage, the log-transformed abundance is modelled by means of a Gaussian distribution with an identity link function. Finally, the predictions of the models in both stages are multiplied at each prediction grid cell in order to obtain an estimation of species abundances.

On a second step, the selected models were used to produce annual predictions of the spatial distribution of the species abundances in a squared mesh grid with cells of 0.01 x 0.01 geographical degrees. Hotspots were defined on an annual basis as those grid cells in which the predicted value of abundance (N/km 2 ) was above a threshold defined as the upper level of the 95% confidence interval of the mean predicted values in the whole time series in the “optimal” depth range for each species and population fraction. This “optimal” was defined as the depth range at which the output of the GAM model for the depth smoother was above 0 (i.e. above the mean abundance when considering the depth effect). The temporal persistence of the hotspots at a grid cell level (probability of being a hotspot) was defined as the amount of years for which a particular cell in the spatial grid was identified as a hotspot divided by all the years in the time series.

Finally, spatial overlap among persistent hotspots of the different species was also investigated. An overlap index, obtained as the sum of species (≥2 species) that had persistent hotspots in a particular grid cell, was calculated. In this case, the overlap was assessed considering four different levels of persistence when defining persistent hotspots: 0.5, 0.6, 0.7 and 0.8.

Fishery dependent data: on board bottom trawlers sampling

This data source was used to provide information on the spatiotemporal distribution and potential main discarding hotspots of the study species. The on-board observers program is carried out within the EU Data Collection Framework (DCF) that takes place in the Balearic Islands from 2003. The monitoring of the bottom trawl fishery consists of five monthly samplings on board randomly selected vessels, which ensures temporal and spatial coverage of the fishing activities.

The information collected by the observers includes the main characteristics of the haul (e.g. date, position, duration, depth) and the weight by species of the commercial catch and discards. Representative size-frequency distributions of the most important species of both the commercial catch and discards are also taken. In the case of crustaceans, these size distributions are collected for males and females separately.

The random sampling scheme used during the onboard sampling does not allow producing maps as in the case of the regular sampling scheme performed during the MEDITS scientific surveys. For this reason, data were analysed for the main commercial fishing grounds of the bottom trawl fleet. VMS data around the Balearic Islands were used to define such fishing grounds (Farriols et al. 2017). The VMS signals were assigned to a net of points defined from a 0.01 degrees resolution grid using Matlab R2013a, and the different fishing grounds were inferred from VMS density contours assigned at each grid point. A total of 32 fishing grounds were identified around the Balearic Islands (Fig. 2). Finally, using expert knowledge of the bottom trawl fishery in the area, each fishing ground was checked in order to differentiate adjacent fishing grounds and delimiting fishing grounds with low densities of VMS. Fig. 2. Main fishing grounds used by bottom trawl fishers from the Balearic Islands.

A total of 1445 fishing hauls performed between 2003 and 2016 were analysed (Fig. 3). The mean species density (number of individuals per hour of effective trawling) and the species size- frequency distribution were computed for each fishing ground. As above, these two parameters were obtained for individuals under and over the MLS and the L50. Both seasonal and interannual variations in species density were analysed.

Fig. 3. Stations of the on board commercial trawlers sampling from the Balearic Islands used in this work.

Feedback with fishers

The main objective of Task 4.3 was “using the knowledge from fisheries scientists in collaboration with fishers to identify the best locations, times and practices to fish to avoid unwanted catch”. Once available, the main results of this Task for the Balearic Islands study area were presented to fishermen in order to check if these results were in accordance with their daily experience at sea. With this goal, bottom trawl fishers with good knowledge of the fishing grounds around the Balearic Islands were consulted.

Discussion

The total list of 19 species previously selected (Table 1), based on the criteria of being under size regulation or having important commercial interest, was analyzed in the case of fishery dependent data. However, the following four species could not be analyzed with fishery independent data owing to limitations in the sampling scheme: the fish Mullus barbatus and the crustaceans Aristeus antennatus, Nephrops norvegicus and Parapenaeus longirostris. In the case of M. barbatus and P. longirostris, two species with size regulation, the low number of individuals taken in the study area prevents analyzing spatiotemporal variations in their populations. The situation is different for A. antennatus and N. norvegicus, two crustacean species abundant in the study area but which undersized population fraction (MLS or L50) are taken in such a small number that it also prevents the analyses.

Fishery independent data: MEDITS surveys

For each species, maps of population density (number of individuals per Km2) and hotspot persistence (in percentage of years) were elaborated for individuals under and over the MLS (if applicable) and the L50. That is, eight maps (4x MLS plus 4x L50) for the species regulated by legal size and four maps (4x L50) for those species without size regulation. A single example of one species of each case is shown in this section: a fish having MLS (hake, Merluccius merluccius) and a skate without MLS (thornback ray, Raja clavata). The maps of all species analyzed with fishery independent data are in Annex 1.

Figure 4 shows the density (above) and persistence (below) maps of hake. Density maps of immature individuals (<32.7 cm) of hake revealed two elongated, narrow areas of high density along the 200 m isobaths in the southwest and northeast (in this case, with some spots at shallower grounds) of Mallorca. Hotspots for mature individuals revealed the same two main

Figure 5 shows the density (above) and persistence (below) maps of the thornback ray. The density maps showed that the species is caught all around the Balearic Islands but at low densities, with the only exception of some areas that changed between juveniles and adults. In the case of juveniles, hotspots were located along the north coast of the Menorca Island and a well defined spot in the west coast of Mallorca. Hotspots of adults extended around Menorca (except the west coast) whereas in Mallorca there is only a spot of slightly higher density than the adjacent areas in the south coast. The persistency maps of the skate showed hotspots for juveniles in two areas of Mallorca (west coast between 50 and 200 m, and south coast around the 200 m isobath) and along the north and south coast of Menorca. Areas of high persistency for adults were found along Menorca (except the west coast) and along the 200 m isobaths of Mallorca (except the southwest coast) and notably in a large area between the 200 and 500 m depth in the south.

The species overlap among the species analyzed in this study considering their corresponding L50 and four levels of persistence (0.5, 0.6, 0.7 and 0.8) are shown in Figure 6. Overlap between two

Outputs

Decision support tools

Task 4.3 was aimed at providing maps of commercial catch operations to allow fishers to make choices of fishing location that avoid discards and provide decision support tools to assist fishers in those choices. In accordance with these aims, the maps and rest of information analysed in this work was presented in a web application developed with Shiny (https://shiny.rstudio.com/), which is an R package to build interactive web apps straight from the software R (https://www.r- project.org/). The Shiny web developed to synthesize these results is structured in different tabs, according to the two data sources used in the analyses: the survey data and the onboard observers’ data. The Survey data tab displays the maps elaborated from data gathered during the MEDITS surveys. It is organized in three sub-tabs, including a general map showing all sampling stations (Map tab) and the maps obtained considering both the L50 (Discards tab) and the MLS (Maturity tab).

The Observer’s data tab, which displays the outputs obtained analyzing data collected onboard commercial vessels, is organized in four sub-tabs. It contains the same three groups as the Survey data tab (Map, Discards and Maturity tabs) together with a new one with the length frequency distribution of the catches by fishing ground and species (Length frequency tab). In this case, however, the Discards and Maturity tabs do not display the information in maps but using graphs of density (number of individuals/effective hour of trawling) of individuals under and above the MLS and the L50. These two tabs are structured in four different sub-tabs showing general results (Discard ratio tab) along with seasonal (Seasonal variability tab) and interannual variations (Inter-annual variability tab) and a table with the data (Summary tab). The Species' Info tab provides a table with the MLS and the L50 for the species included in the analysis. It also provides links to other websites with general information on the biology and ecology of the species.

Finally, the About tab contains general information about the DiscardLess project and the development of this Shiny app. In accordance with the main aim of providing decision support tools to assist fishers to make choices of fishing location to avoid discards, the following Shiny app, which contains all the results obtained in this work, were developed: https://lucia2r.shinyapps.io/prueba_nolog/. Figure 13 is a snapshot of the home page displaying the different tabs described above: Observer’s data, Survey Data, Species’ Info and About. In this example, the snapshot shows the specific case of the Observer’s data tab, which contains the following sub-tabs (also available for the Survey Data tab): Map, Length Frequency, Discards and Maturity. On the left hand side there appear the dropdowns to choose the species and the fishing grounds to be shown. As it is shown below the dropdowns, only those fishing grounds containing a number of sampling stations higher than 25 were used in the analyses. Figure 14 shows an example of a species that does not have MLS, in this case the common octopus (Octopus vulgaris) taken by commercial trawlers (Observer’s data tab). The snapshot shows the population structure of this species for all fishing grounds where the species was caught (sample size >25) under the Length Frequency tab. As this species does not have MLS, the Discards tab does not contain graphs but only a message informing about this (Fig. 15). In this case, only graphs considering the MLS can be produced showing the information for immature and mature individuals: total (Fig. 16), seasonal (Fig. 17) and interannual (Fig. 18) variability in species density (number of individuals per hour of trawling). The Species' Info tab (Fig. 19) provides a table with the MLS and the L50 for the species included in this work. It also provides links to other websites with general information on the biology and ecology of the species. Finally, the About tab contains general information about the DiscardLess project and the development of this Shiny app (Fig. 20).