Environmental preferences of tuna and non-tuna species associated with drifting fish aggregating devices (DFADs) in the Atlantic Ocean, ascertained through fishers’ echo-sounder buoys

Author’s Accepted Manuscript Environmental preferences of tuna and non-tuna species associated with drifting fish aggregating devices (DFADs) in the Atlantic Ocean, ascertained through fishers’ echo-sounder buoys Jon Lopez, Gala Moreno, Cleridy Lennert-Cody, Mark Maunder, Igor Sancristobal, Ainhoa Caballero, Laurent Dagorn

PII: DOI: Reference:

www.elsevier.com/locate/dsr2

S0967-0645(17)30040-1 http://dx.doi.org/10.1016/j.dsr2.2017.02.007 DSRII4200

To appear in: Deep-Sea Research Part II Cite this article as: Jon Lopez, Gala Moreno, Cleridy Lennert-Cody, Mark Maunder, Igor Sancristobal, Ainhoa Caballero and Laurent Dagorn, Environmental preferences of tuna and non-tuna species associated with drifting fish aggregating devices (DFADs) in the Atlantic Ocean, ascertained through fishers’ echo-sounder buoys, Deep-Sea Research Part II, http://dx.doi.org/10.1016/j.dsr2.2017.02.007 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Environmental preferences of tuna and non-tuna species associated with drifting fish aggregating devices (DFADs) in the Atlantic Ocean, ascertained through fishers’ echo-sounder buoys Jon Lopez1*, Gala Moreno2,1, Cleridy Lennert-Cody3, Mark Maunder3, Igor Sancristobal1, Ainhoa Caballero1, Laurent Dagorn4 1

Azti-Tecnalia. Herrera kaia, portualdea z/g, 20110, Pasaia, Spain.

2

International Seafood Sustainability Foundation (ISSF), 805 15th Street NW,

Washington, DC 20005, USA 3

Inter-American Tropical Tuna Commission, 8901 La Jolla Shores Drive, La Jolla, CA

92037, USA 4

Institut de Recherche pour le Développement, IRD, UMR EME 212, Avenue Jean

Monnet, CS 30171, 34203 Sète Cedex, France *

corresponding author: Tel.: +34 634 209 738; fax: +34 9465472555. [email protected]

Abstract Understanding the relationship between environmental variables and pelagic species concentrations and dynamics is helpful to improve fishery management, especially in a changing environment. Drifting fish aggregating device (DFAD)-associated tuna and non-tuna biomass data from the fishers’ echo-sounder buoys operating in the Atlantic Ocean have been modelled as functions of oceanographic (Sea Surface Temperature, Chlorophyll-a, Salinity, Sea Level Anomaly, Thermocline depth and gradient, Geostrophic current, Total Current, Depth) and DFAD variables (DFAD speed, bearing 1

and soak time) using Generalized Additive Mixed Models (GAMMs). Biological interaction (presence of non-tuna species at DFADs) was also included in the tuna model, and found to be significant at this time scale. All variables were included in the analyses but only some of them were highly significant, and variable significance differed among fish groups. In general, most of the fish biomass distribution was explained by the ocean productivity and DFAD-variables. Indeed, this study revealed different environmental preferences for tunas and non-tuna species and suggested the existence of active habitat selection. This improved assessment of environmental and DFAD effects on tuna and non-tuna catchability in the purse seine tuna fishery will contribute to transfer of better scientific advice to regional tuna commissions for the management and conservation of exploited resources. Keywords: FAD; environmental preferences; habitat; GAMM; purse seine; echosounder buoy; tuna

Introduction Organisms are heterogeneously distributed in the marine environment. They are frequently connected with particular habitats and their requirements can change through their ontogeny. Indeed, biological processes are synchronized with environmental processes within specific sets of temporal and spatial scales (Bakun, 1996). This is well illustrated in the pelagic realm, and particularly in top predators’ communities such as tropical tunas and accompanying species (Tew Kai and Marsac, 2010). Thus, the aggregative response of pelagic populations has to be investigated at suitable spatialtemporal scales. Understanding the mesoscale features involved in the dynamics of

2

pelagic populations is helpful to elucidate the importance of a number of processes influencing their behavioural patterns. Floating objects on the surface of the tropical and subtropical oceans tend to aggregate a number of marine species, including commercially valuable tropical tunas and some other accompanying non-target species (Castro et al., 2002). Taking advantage of this evolutionary associative behaviour, tropical tuna purse seiners annually construct, deploy, and monitor an estimated 50,000-100,000 drifting fish aggregating devices (DFADs) worldwide (Baske et al., 2012; Maufroy et al., 2016; Scott and Lopez, 2014; Ushioda, 2015), many of them equipped with satellite linked echo-sounder buoys, which provide fishers with remotely collected rough estimates of DFAD-associated fish biomass as well as accurate geolocation information along the trajectory (Lopez et al., 2014). However, the exact reasons driving the aggregative behaviour around DFADs remain elusive, as well as most of the factors motivating the activities (i.e. excursions, residence and absence times, etc.) related to the fine and large scale movements of DFAD-associated species. Today, nearly half of the biomass of the world market tropical tunas (i.e. skipjack Katsuwonus pelamis, yellowfin Thunnus albacares, bigeye T. obesus) are fished under DFADs (Fonteneau et al., 2013). The increasing use of floating objects by purse seine fisheries has led to international concern because it raises the possibility of three potential impacts: (i) reduction in yield per recruit of some target species (i.e. yellowfin and bigeye), (ii) increased by-catch and perturbation of the pelagic ecosystem balance and (iii) deleterious alteration of the normal movements of the species associated with DFADs (i.e. Ecological Trap (Hallier and Gaertner, 2008; Marsac et al., 2001)) (Dagorn

3

et al., 2012b). In spite of this concern, the effects of the massive deployment of DFADs on the pelagic ecosystem and on individual and collective fish behaviour, ecology, and biology are still unknown (Anonymous, 2014). However, tuna regional fishery management organizations (t-RFMOs) are constantly exploring ways, and have already adopted conservation measures, to try to minimize the impact of these devices in the ecosystem, such as establishing time-area closures (Torres-Irineo et al., 2011) or promoting the use of non-entangling DFADs (Goñi et al., 2015; Hernández-García et al., 2014), among others. Determining which factors can explain the presence or biomass density of tunas and other species at DFADs would be helpful to improve management of the exploited resources, especially in a changing ocean. Nevertheless, many questions remain unresolved: are tunas and other species associated to DFADs only under certain environmental conditions? Do pelagic species show preferences for DFADs with particular trajectories or entrained in areas with specific oceanographic characteristics? Which are the factors motivating tunas and non-tuna species to join or leave the DFAD? Are these factors specifically hierarchized? Up to now most of the studies analysing the environmental preferences of tropical tuna species have been based on catch data, especially longline CPUE (Catch Per Unit of Effort) data and very rarely purse seiner CPUE data (Table 1). Using fishery data to investigate ecological issues can lead to some misleading conclusions (Maunder et al., 2006) as catch information are biased and limited by fishing effort, sampling coverage, vessels’ technology and capacity, and geographical constraints. Some fishery-independent data are already being collected and used by the scientific community to investigate the behaviour of fish associated with DFADs (i.e. scientific acoustic surveys (Moreno et al., 2007b), underwater visual 4

surveys (Taquet et al., 2007), and acoustic telemetry (Forget et al., 2015; Matsumoto et al., 2016; Matsumoto et al., 2014; Schaefer and Fuller, 2013)), but most analyses to date have been based on individual behaviour or low sample size, hindering inference on collective behaviour patterns and abundance dynamic responses. Thus, there is still a need for additional and complementary use and implementation of new methodologies and data sources that allow investigating collective behaviour of DFAD-associated fish at large scale. In this paper, we investigate the relationship between environmental variables and DFAD-associated biomass derived from fishers’ echo-sounder buoy data. Echo-sounder buoy data are fishery-dependent, but are not directly related to the catch, and as such are less dependent on the fishery than CPUE, which is the most commonly used fisheriesdependent data type. Generalized Additive Mixed-effect Models (GAMMs; (Bates, 2010; Grafarend, 2006)) have been used to adequately monitor and measure the response of each fish species to certain environmental, biotic and abiotic changes, which will provide new knowledge on the effect of DFADs on the pelagic ecosystem, assisting the development of potential conservation measures and management decisions.

Material and Methods Data collection

A series of fish biomass data from 76 Satlink echo-sounder buoys (Satlink, Spain, www.satlink.es) attached to DFADs were gathered between May 2009 and May 2012 in the central and eastern tropical Atlantic Ocean through the ECOFAD Program established in collaboration with the ANABAC ship-owner association. Detailed 5

information on the Satlink echo-sounder buoy can be found in Robert et al. (2013) and Lopez et al. (2016). In the present study, all the used buoys spent at least 5 days at sea; it is assumed that shorter soak times would not be representative of processes affecting DFAD. About 72% of the echo-sounder buoys were attached to newly deployed DFADs (i.e. virgin DFADs), which allowed to account for the effect of soak time (i.e. time spent in the water since initial deployment) in the models. Echo-sounder data were obtained from acoustic samples at sunrise (i.e. the time of the day in which fish are supposed to be more closely aggregated under the DFAD (Josse et al., 2000; Moreno et al., 2007a)).

Echo-sounder buoy data were analysed following the model proposed by Lopez et al. (2016) for Satlink buoys in the Atlantic ocean, where knowledge of the vertical distribution of DFAD-associated fish aggregations obtained through scientific experiments around DFADs (Forget et al., 2015; Moreno et al., 2007b) was used to set virtual limits between fish categories. Thus, the sa (area backscattering coefficient, m2 m-2; Maclennan et al. (2002)) from the first 25 m were assigned to non-tuna species, whereas the sa from 25 m to 115 m were assumed to correspond to tunas. Similar depth limits were adopted in previous studies using the same echo-sounder buoy to separate non-tuna species from tunas (Lopez et al., 2010; Robert et al., 2013). Area backscattering coefficients for each layer were transformed into biomass estimates (in metric tons, t) using a depth layer echo integration procedure (Simmonds and MacLennan, 2005), based on i) the target strengths suggested by Doray et al. (2007) and Moreno et al. (2007b), for non-tuna and tuna species aggregations (non-tunas -42 dB; tunas -35.1 dB), respectively; and ii) the weight of the most common sizes of fish found at FADs (i.e. 1 kg. for non-tuna species (F. Forget, pers. comm.); 2 kg. for tuna, 6

respectively (Floch et al., 2012); see Lopez et al. (2016) for details).). The estimated fish biomasses were assumed to represent the DFAD-associated biomass for the different fish groups. Positive biomass of fish groups were log-transformed (Loge(Biomass)) to assure normality (Zarauz et al., 2008).

For each acoustic record, the following oceanographic data were used: temperature at the sea surface, at 30 m, at 75 m and at 100 m depth (further referred to as SST, T30, T75 and T100; in °C), salinity at 30 m, 75 m and 100 m depth (S30, S75, and S100; in PSU), chlorophyll-a concentration the same day of the sampling event and 18 days before

(Chl-a

and Chl-a18, respectively; in mg C m-3), sea-level anomaly (departures of the sea surface height from long term mean, SLA; in cm), depth (bathymetry, D; in m), geostrophic current (WG; in kn), total current (WT; in kn), mixed layer depth (TD; in m), and thermocline gradient (TG; in °C). Oceanographic information for the region and time period of this study was provided by satellite data obtained from different combinations of

AVHRR

(SST;

4

km

resolution),

ERS-2,

Topex/Poseidon,

Jason-1/2,

ENVISAT/GFO, CRYOSAT (SLA and WG; 25 km resolution) and MODIS-MERIS (Chl-a, Chl-a18; 4 km resolution) sensors and outputs from ocean models (T30, T75, T100, S30, S75, S100, WT, TD, TG; 25 km resolution). This information was processed and provided by the CLS (Collecte Localisation Satellite, France, https://www.cls.fr). Depth values were taken from GEBCO gridded bathymetry data (British Oceanographic Data Centre, UK, www.gebco.net). DFAD-variables such as bearing (in °), speed (in kn), and soak time (in days) were also used as explanatory variables. Total amount of non-tuna species (BC; in t) was also included in the models as a proxy for biological interaction, defined on the basis of the most logical trophic level hierarchy and 7

representing the potential ecological relationships (i.e. attraction power, etc.) existing between fish groups: the effect of non-tuna species was considered in tuna models, but not the other way. After data processing, the final dataset used in this study consisted of 3671 acoustic samples (Fig. 1).

Methods

As a preliminary exploration of the relative effect of covariates on the dependent variable (i.e. fish group biomass) univariate generalized additive models (GAMs; Hastie and Tibshirani (1990)) were established using the package mgcv (Wood, 2014) in the statistical software R (Team 2013). Similarly, predictor covariates were examined for correlation using pair plots and Pearson’s rank correlations. To avoid correlation, one covariate from covariate pairs with a correlation > 0.7 and < -0.7 was removed from the variable selection process (Dormann et al., 2013; Hassrick et al., 2016). As an additional measure to avoid collinearity, a variance inflation factor analysis (VIF), from the package usdm (Naimi, 2015) was conducted using a cut-off value of 3 (Zuur et al., 2009b). Based on these analyses, T30, T75, S30, S75, S100 and TD were removed owing to correlation/collinearity with more ecologically important variables. All other covariates available for model selection had low cross-correlation and cross-collinearity scores. In addition, a preliminary analysis with a non-parametric spline correlogram (BjØrnstad and Falck, 2001) from the package ncf (Bjornstad, 2009) suggested that both biomass data for non-tuna species and tuna lack spatial autocorrelation, and as such, the

8

inclusion of spatial terms in the models was considered unnecessary (Louzao et al., 2009). A two-part delta-lognormal model (e.g. Lo et al. (1992); first part: presence and absence of fish group biomass; second part: biomass density, given its presence) was developed for the DFAD-associated biomass data of each fish group to explore the effects of environmental factors on fish distributions. A GAMM with a binomial error distribution and a logistic link function was used to model the presence and absence of biomass. A GAMM with a Gaussian error distribution and identity link function was used to model the biomass density. In the latter case, the natural logarithm of positive biomass (i.e. log (Biomass)) was used as the dependent variable. All remaining environmental covariates were considered initially for both parts of the model. Each GAMM was fitted using: (i) thin plate regression splines to model nonlinear covariate effects, except for DFAD bearing effect, where a cyclic cubic regression spline was used to reflect the cyclical behaviour (Wood, 2006), , (ii) a DFAD-specific random effect (i.e. Buoy ID) to account for the autocorrelation structure of the data set, and (iii) the dimension, k, representing maximum degrees of freedom of each smooth, was manually limited by k = 4 to avoid excessive flexibility and model overfitting and to simplify the interpretation of the results (Cardinale et al., 2009; Giannoulaki et al., 2013; Jones et al., 2014). Smoothing parameters were estimated using the Laplace (ML) and REML methods for the binomial and the Gaussian error distributions, respectively (Grafarend, 2006). The following models were fitted for each fish group: Response variable ~ s(Ca k=4) + s(Cb k=4) + ... + s(Cz k=4) + c(Bearing, k=4), random=~(1|Buoy)

9

where Response variable is either the probability of presence of a fish group or the corresponding biomass density considering only the positive observations, s represents a penalized thin plate regression spline type smoother for environmental covariates (C), c specifies a cyclic penalized cubic regression spline smoother, k is the maximum degrees of freedom allowed to the smoothing function, and random = ~(1|Buoy) is an ad hoc way of accounting for the autocorrelation structure of the data set in GAMMs. When the estimated degrees of freedom of the splines of the explanatory variables were close to their lower limit (i.e. close to 1 for univariate smoothing), the covariate in question was considered to be linear (Wood, 2008). The selection of the effective covariates to include in each GAMM was performed applying a stepwise elimination, based on the method proposed by Wood and Augustin (2002). Covariates were deleted from the models following two criteria: (i) the approximate 95% confidence band for the smooth term included zero throughout; and (ii) the AIC (Akaike Information Criterion; Akaike (1974)) score decreased when the term was dropped. Models with the lowest AIC scores were selected. All the GAMM models were fitted using the gamm4 package for R (Wood, 2011). Because there is no generally accepted procedure for estimating model fit for GAMMs (Wood, 2006; Zuur et al., 2009a), this study employed the approach used by Gilman et al. (2012) of fitting an equivalent GAM to derive the percent deviance explained, and to evaluate the importance of explicitly accounting for DFAD-specific heterogeneity within the GAMMs. This approach is based on fitting a GAM to the data (with the same terms as the GAMM, except without the random effects), then fitting linear models to the GAM deviance residuals, with and without the random effect terms,

10

and using AIC and likelihood ratio tests (Wood, 2006) to evaluate the model improvement obtained from inclusion of the random effects. Given that the number of buoys (i.e. DFADs) used in the study was not extremely large, GAMM equivalent GAMs with “Buoy” as a random effect type spline were also developed (hereafter called GAMre) to obtain a second rough approximation of the potential percent deviance explained by the GAMMs.

Results

Table 2 shows AIC values for the GAMM and equivalent GAM, and results of a log-likelihood ratio test. In general, the GAMMs were significantly better fitting models than the equivalent GAMs, which did not account for the autocorrelation structure of the data set. With the exception of the models for biomass density for non-tuna species, the log-likelihood ratio test for the DFAD-specific random effect variance yielded a pvalue <0.0001. Similarly, the GAMre also showed better AIC and deviance scores than the equivalent GAMs. The results of the GAMM models are shown in Table 3 and in Figures 2-5 as plots of the best-fitting smooths for the conditional effect of the covariates on the parameter of interest (i.e. DFAD-associated biomass of a fish group). The models produced suitable fits to the data set based on the percent deviance explained from the equivalent GAMs and GAMre, with no apparent aberrant residual behaviour.

Non-tuna species

11

The logistic GAMM for the presence and absence of non-tuna species explained at least 14.9% of the deviance (28.7% by the GAMre) with an adjusted r2 of 0.13 (Table 23). There were five variables that met the significance criteria outline in the Methods section, ordered according to variable significance (Table 3, Fig. 2): SST, DFAD soak time, DFAD bearing, Chl-a, and DFAD speed. The probability of non-tuna biomass was higher at relatively low values of SST (24-26 °C), as well as at moderate values of Chl-a, soak times and DFAD speeds. Additionally, DFADs heading in a westward direction were more likely to have nontuna species. In contrast to the results for the presence and absence logistic model, there appeared to be little information in the data related to the amount of biomass of non-tuna species. The GAMM for biomass density of non-tuna species explained the lowest percentage of the deviance among all fish group models (2.35%; 5.37% according to GAMre), with a r2 of 0.014 (Table 2-3). The variables selected to build the model were, in order of variable significance (Table 3, Fig. 3): SLA, SST, and DFAD speed. Positive values of SLA were associated with higher non-tuna biomass, and a decreasing linear relationship with SST and DFAD speed was found.

Tunas The GAMM for tuna presence and absence explained at least 8.66% of the deviance (23.6% according to the GAMre), with a r2 of 0.06 (Table 2-3). The seven selected

12

covariates were, in order of variable significance (Table 3, Fig. 4): DFAD soak time, SLA, WG, WT, SST, DFAD bearing, and Chl-a. The probability of tunas was generally higher at high values of DFAD soak time, except at around 100-150 days, where a flat trend was observed, and at positive values of SLA. Higher probability of presence occurred at low WG and intermediate WT values (extreme values also showed higher probability of tuna presence although the confidence intervals are too wide at this level). The highest values of the probability of tunas occurred at the intermediate values of SST (24-28 °C) and DFADs heading in southwest and northwest directions, while the highest Chl-a values were also associated with higher probability of tunas. The GAMM for the biomass density of tunas explained at least 5.17% of the deviance (29.2% according to the GAMre) and had an r2 of 0.02 (Table 2-3). Six covariates were selected for this model, in order of variable significance (Table 3, Fig. 5): DFAD bearing, DFAD soak time, TG, T100, BC, and D. Higher biomass was associated with DFADs heading in west directions and low values of soak time. An increase in biomass was detected for low values of TG (< 2 °C) and high values of T100. Tuna biomass was also higher at intermediate BC values, while lower biomass was associated with the highest values of depth.

Discussion

13

This is the first time that the relationships between DFAD-associated fish biomass and environmental factors have been described in repeated large-scale non-invasive samplings, giving an overall description of the relative effect of the different variables on the distribution of DFAD-associated non-tuna species and tunas. Acoustic sampling was undertaken at a significant spatial-temporal resolution (Fig. 1), covering three consecutive years (2009-2012) and an area larger than the typical DFAD fishing grounds. This data set was less affected by seasonality, effort, and other typical limitations of fisheries data, which provided several obvious advantages compared to those studies using catch data. The results indicated that echo-sounder buoy technology, combined with remote sensing data and improved quantitative techniques, has the potential to enable remote sampling and investigation of DFAD-associated fish aggregations and their environment. The GAMM method was shown to be a useful tool to use in describing the interactions between DFAD-associated fish biomass and physical and biological variables, in order to produce models with a good descriptive capability. All four models used included a combination of hydrographic and DFAD (DFAD speed, bearing, and soak time) terms. With the exception of the model for non-tuna biomass density, probably due to the fact that non-tuna species biomass at DFADs shows low variability (Dagorn et al., 2012a), the other presence and absence and biomass density models accounted for a reasonable amount of the variability in the data. Moreover, being able to sample different fish groups at the same spatial-temporal resolution as the biotic and abiotic variables is helpful for a better understanding of the factors affecting DFAD-associated fish dynamics. It permits the assessment of the potential relative

14

importance of factors on fish presence and biomass density, as well as allows the evaluation of possible DFAD fish retention and departure mechanisms.

Effects of oceanographic variables on DFAD associated fish In the present study, the contribution of the oceanographic and DFAD variables was significant for all the models. However, the type of relationship between environmental variables and biomass differed for the fish group considered. This suggests that, at this scale, the heterogeneous distribution of non-tuna species and tunas may be determined by different preferred habitats. In the tropical Atlantic Ocean, the most significant high productivity features include both persistent and seasonal structures, such as the permanent Mauritanian and Angolan upwellings and occasional emergence events located along the Senegalese, Ghanaian, and Gabonese coasts and equatorial divergence in boreal summer (i.e. cold season, June to September) (Fonteneau and Marcille, 1993; Longhurst, 2006; Longhurst and Pauly, 1987; McGlade et al., 2002). (Fig. 6). Model results highlighted the importance of productivity for non-tuna and tuna species distribution. Indeed, productivity has always been relevant for the distribution of marine species, especially for large predators (Longhurst and Pauly, 1987). In that sense, Chl-a plays a key role in marine ecosystems as it is the source of the energy circulating through the trophic levels, and thus, it can be seen as a proxy of prey enrichment (Marsac, 2013). Fonteneau et al. (2008) found interesting relationships between circumstantial Chl-a peak levels 18 days before and free swimming tuna abundance in the Indian Ocean. The present study, which focuses on the Atlantic Ocean and DFAD-associated fish, did not find any 15

relationship between this variable and tuna distributions. It is worth mentioning that the exceptional Chl-a values found by Fonteneau et al. (2008) were of more than 60 mg C m-3 while the maximum values in the present data set were only about 4 mg C m-3. Despite this, this study reveals significant correlations between Chl-a for the same day of observations and both presence of tunas and non-tuna species (note that for non-tunas the probability was higher at moderate values of Chl-a). This relationship is widely known by fishers and scientists (Chassot et al., 2011; Fraile et al., 2010; Marsac, 2013; Song et al., 2008), who are now increasingly using remote sensing and satellite imagery to identify phytoplankton blooms potentially associated with tuna hotspots. This highlights the current value of employing near real-time cutting-edge on board oceanographic maps and understanding the effect of some environmental factors that enable predicting the distribution of tuna resources for both fishing and management purposes. The correlation between SLA and productivity, especially in the tropics, has been observed previously (Benitez-Nelson et al., 2007; Lopez-Calderon et al., 2006; Tew Kai and Marsac, 2010; Wilson and Adamec, 2001). The present work has proven SLA to be significantly related with both tunas and non-tunas species distribution (i.e. SLA >0) (Figs. 3 and 4), which is in accordance with previous findings (Tew Kai and Marsac, 2010), and suggests that tuna and non-tuna species may favour anomaly areas to take advantage of the higher surface primary production. According to the literature, both positive and negative SLA values may be related to eddies leading to enhanced ecosystem productivity. However, the interactions between SLA and eddy dynamics are very complex and involve a number of mechanisms that are not well understood (Kahru

16

et al., 2007). Further studies should analyse in detail whether our anomalies are linked to eddies or other type of mesoscale features. Water temperature has always been linked with productivity as well, cold waters typically holding more nutrients than warm waters (Longhurst, 2006). Most tunas tend to concentrate along thermal discontinuities such as oceanic fronts (Fiedler and Bernard, 1987; Mugo et al., 2014; Sund et al., 1981), fish principally staying in the warm part and moving to the cold part to feed. Several authors have linked warm waters with higher catches of large tuna (Lan et al., 2011; Stretta, 1993; Zagaglia et al., 2004) while others found the opposite (Farrell et al., 2014; Lan et al., 2012; Song et al., 2009; Song et al., 2008). Results of the current research support the idea of a negative correlation between SST and non-tuna species abundance, indicating that these individuals may show predilection by cold high productivity waters. However, results suggested that tunas preferred moderate temperature values in the surface (24-28 ºC) and relatively warm deep waters. These signs, in conjunction with the negative relationship found with the thermocline gradient, may indicate that tunas show preference for well mixed waters that favour their vertical habitat extension and movements. In fact, tunas are well known to regularly conduct excursions to deep waters to feed on the deep scattering layer (Bertrand et al., 2002a; Dagorn et al., 2006). Ideally, specific studies should couple tropical tuna tagging data with environmental data sets and biologging (i.e. individual biological factors that may provide information on the condition of the fish) to try to infer behavioural characteristics in relation to physiological states and different oceanographic conditions in the water column. Results would provide interesting evidence to better understand catchability of the fish and interpret buoy data.

17

Tunas also showed higher association rates with DFADs in relatively shallow areas (i.e. continental shelf break, seamounts) (Fig. 5), where large pelagics are well known to habit principally due to foraging advantages, and possibly reproductive and navigational benefits (Allain et al., 2006; Dubroca et al., 2013; Fonteneau, 1991; Holland and Grubbs, 2008; Morato, 2006). Fishers are aware of this relationship and as such, certain supply vessels are anchored around seamounts in particular areas of the Atlantic Ocean. This interesting relationship should be further investigated by analysing catch rates of species and data from acoustic surveys (either buoy data or scientific cruises) in relation to the distance to underwater elevations to better understand the effective area of these structures. Tuna environmental preferences observed in the present work are in accordance with most of the published research investigating those relationships with, primarily, non-FAD-associated catch (i.e. which are not supposed to be affected by the Ecological trap) or experimental data (Table 1). Furthermore, higher individual concentrations, and hence larger schools, have been found in favourable habitats. As Roger (1994) indicated, tunas adopt a nomadic strategy in small schools when food resources are scarce and form large schools when they are abundant. This suggests that, as many authors have previously stated (Bertrand et al., 2002b; Doray et al., 2009; Sund et al., 1981), larger tuna aggregations are usually located in the preferred mixed layer of actively selected beneficial locations, likely during seasonal migratory movements. The results from the present work support that idea and suggest that the hypothesis of tuna being trapped by DFADs in a non-favourable environment might be incorrect. Nevertheless, further biological and ecological complementary studies would be desirable to test this hypothesis, as this information would be helpful to assist an 18

adequate sustainable management and assessment of exploited stocks of tropical tuna and non-tuna species.

Effects of DFAD-variables on DFAD-associated fish Results of the GAMM models showed higher non-tuna and tuna abundances for DFADs heading westward. Different reasons may be behind this correlation. According to

Morrow

et

al.

(2004),

eddies

propagate

westward

and

equatorward

(anticyclonic)/poleward (cyclonic). As Doray et al. (2009) suggested, tuna may follow these structures across their migration paths, which also appear to be westward in the tropical Atlantic Ocean. Although tropical tunas undertake migrations of different scales in this region (Anonymous, 2004; Bard et al., 1993; Delgado de Molina et al., 2005), the currently accepted stock structure of tunas in the Atlantic Ocean (Bard and Hervé, 1994; Bard et al., 1993; Fonteneau and Soubrier, 1996; Zagaglia et al., 2004) suggests that juveniles of yellowfin and bigeye, as well as skipjack of any sizes, move along the west African coast, NE South America and central equatorial zone from the main recruitment area in the Gulf of Guinea (i.e. westwards), likely in mixed schools (Bard et al., 1993). Conversely, adults of yellowfin may extend east at the beginning of the year and migrate westward around the second/third quarter (Lan et al., 2011; Zagaglia et al., 2004). Very little tagging information is available so far for pelagic species in the Atlantic Ocean. However, the ICCAT (International Council for the Conservation of Atlantic Tunas) has recently launched a large-scale tuna tagging program, similar to that conducted years ago in the Indian Ocean (Murua et al., 2015). This experiment will provide high quality movement data, necessary to develop current hypotheses and

19

improve scientific knowledge on the ecology and biology of the species and assure sustainability. Further works should investigate this bearing-biomass interaction in deep, as it could be possible that tunas and non-tuna species associate to DFADs while following their general migration paths. Besides DFAD bearing, speed of the DFAD, as well as more general current speeds (WG, WT), have also proven to be significant and negatively correlated with tuna and non-tuna species, which is in agreement with previous findings in the field (Kakuma, 2000). In general, both fish groups showed predilection by speeds around 0.5-1 kn, or lower. The energetic cost of being associated with relatively fast-DFADs (> 1-1.5 kn) may be negative, and in this scenario, non-tuna species and tunas could be forced to leave the floating object. Similar energetic difficulties are experienced by small tunas when associating with dolphins in the Eastern Pacific Ocean, where large tunas, in turn, successfully accomplish to swim with them (Edwards, 1992). These findings are in accordance with the information provided by fishers during interviews (J. Lopez, pers. comm.), who stated that the presence of individuals at DFADs could be related to the dominant trajectory vector values (bearing and speed) of a floating object. The way in which fishers build their own artificial DFADs, bamboo rafts with underwater nets fitted with weights in the form of an anchor to decrease DFAD speed, support this interpretation. Interestingly, the speed and direction of floating objects should be noted by fisheries observers on board and compared with the current results, as tunas may primarily associate with DFADs that propagate with certain trajectory values. These data could also help differentiate among potential hypotheses on the attraction-retention power of floating objects.

20

Whereas there seems to be a general positive correlation between soak time and probability of fish at DFADs, especially up to values around 100 days (i.e. non-tuna species) or more (i.e. tunas), tuna abundance at DFADs is not positively correlated with soak time, indicating a complex colonization process not well understood yet. Different reasons have been identified to potentially explain this variable relationship: it can be hypothesized that DFADs may lose biomass if the original objects’ structures begin to fail (i.e. damaged rafts, sinking of underwater parts, etc.) or a complex ecological succession leads large top predators to the DFAD (i.e. marine mammals, marlins, etc.). According to fishers, the average lifespan of an artificial DFAD is about 5-12 months in the Atlantic Ocean. Fishers also state that the presence of marine mammals can negatively impact other species occurrence at DFADs (J. Lopez, pers. comm.), which is also in accordance with previous findings by Japanese and French scientists (Brehmer et al., 2011; Ohta and Kakuma, 2005). It may also happen that the association of tunas is opportunistic and partially independent of the ecological succession or maturation process of the floating object and that other factors are involved in the attraction power of the DFAD. This study found correlation between the abundance of non-tuna species at DFADs and tuna abundances, with a preference around 1 t of non-tuna species (i.e. note that the average size of bycatch at DFADs by European purse seine fleet in the Atlantic Ocean is 0.8 t; Amandé, pers. comm.). This positive relationship has also been found in studies investigating the behaviour of aggregations at DFADs for intermediate or small time scales (Lopez, 2015; Robert et al., 2014; Robert et al., 2013). The presence of non-tuna species at DFADs may facilitate tuna to locate and associate with floating objects. In fact, sound produced by other animals (Ghazali et al., 2013), as well as other sensory signals (chemicals, etc.) (Dempster and Kingsford, 2003) have been

21

suggested as potential attraction and orientation cues for relocating FADs after excursions. Further studies should investigate this soak time-biomass interaction in detail as understanding the colonization process of a floating object would provide interesting knowledge to be applied in the management of DFADs and conservation of exploited species.

Other factors potentially affecting tuna distribution Although we found significant relationships between specific environmental variables and fish distribution, the consideration of new variables, such as SST and Chla gradients at different spatial-temporal scales (e.g. days, weeks), EkE (eddy kinetic energy), number and distances to fronts in an area, oxygen concentration, large scale oscillation indices, etc. would possibly provide complementary information to current models. Several authors have noted the crucial effect of oxygen availability on pelagic species distributions, especially for large predators (Potier et al., 2014; Prince et al., 2010; Stramma et al., 2008). In fact, the available habitat of certain tropical species may be reduced due to the expansion of the oxygen minimum zone (Stramma et al., 2012), which potentially could affect tuna and non-tuna species geographical distributions. Information on oxygen levels could not be included in the GAMMs of this study because we did not have access to adequate fine-scale measurements for our study period and location. Future studies should consider surface and subsurface oxygen concentration values in their models to integrate the effect of key limiting factors on the distribution of both tunas and non-tuna species associated with DFADs. 22

The relationships between tropical tuna CPUE and large-scale climate indices (Southern Oscillation Index in the Pacific Ocean, the North Atlantic Oscillation in the Atlantic Ocean, and the Indian Ocean Dipole in the Indian Ocean) have been demonstrated to be significant by several authors (Lan et al., 2012; Ménard et al., 2007; Rouyer et al., 2008; Santiago, 1998; Stenseth et al., 2002). The limited time series of echo-sounder buoy data (3 years) compared to the periodicity of the NAO (~4 years) means that the relationship cannot be studied with our data set. Longer time series will be necessary for exploring the robustness of the potential interaction.

Conclusion

This is the first time that an overall description of the relative effect of the different variables on the distribution of DFAD-associated non-tuna species and tunas has been provided using repeated large-scale non-invasive samplings. Echo-sounder buoy data, less affected by the typical limitations of fisheries data (e.g. seasonality, effort), combined with remote sensing data and improved quantitative techniques, showed the potential to investigate aspects of the ecology and behaviour of DFAD-associated fish. Despite that thousands of DFADs are deployed annually worldwide, the flow of echosounder buoy data from vessel owners to research institutes is still scarce due to confidentiality and sensitive aspects. A data exchange system that preserves the privacy and ensures future bilateral advantages could help in effective fishers-science collaboration. The large amount of data continuously collected by echo-sounder buoys would assist to promote a large variety of studies of interest for the scientific 23

community and managers. For example, these data would help to better understand the different aspects involved in the collective behavior of aggregations at DFADs at different spatial-temporal scales but also to progress towards alternative abundance indices of tunas and accompanying species (i.e. acoustic-based indices or improved purse seine CPUEs by accounting for catchability or behavioral patterns). Initiatives like this work would be helpful to better understand, manage and assess the potential effects of FADs on the pelagic ecosystem, advancing toward the long-term sustainability of the tuna fisheries in a changing ocean.

Acknowledgments

The authors would like to sincerely thank Spanish fishing masters of tuna purse seiners in the Atlantic Ocean who kindly agreed to deploy the echo-sounder buoys used in the present study. We would like to thank the purse seine company owners and the association that represent them, ANABAC. Also, sincere thanks to CLS (Collecte Localisation Satellite) for providing the oceanography data. We sincerely thank Dr. Jerry Scott for revising the English. This study was a part of the European project MADE (Mitigating Adverse Ecological Impacts of open ocean fisheries; funded by DG Research, collaborative project n° 210496) and ECOFAD programme (a project funded by ANABAC). This study was partly funded by a PhD grant by the Fundación Centros Tecnologicos Iñaki Goenaga to Jon Lopez. This paper is contribution number xxx from AZTI-Tecnalia.

24

References

Akaike, H., 1974. A new look at the statistical model identification. Automatic Control, IEEE Transactions on 19, 716-723. Allain, V., Kirby, D.S., Kerandel, J.A., 2006. Seamount Research Planning Workshop Report, 20-21 March 2006 [EB IP-5]. Kolonia, Pohnpei State, Federated States of Micronesia: Western and Central Pacific Fisheries Commission. Scientific Committee Second Regular Session, Manilla, Philippines, 7-18 August 2006, 2nd. 56 p. Anonymous, 2004. Sesión de evaluación ICCAT 2003 del stock de rabil (Mérida, México, 21 – 26 de julio). Collective Volume of Scientific Papers, ICCAT, 56: 443 – 527. Anonymous, 2014. Report of the ISSF workshop on FADs as ecological traps. ISSF Technical

Report

2014-03.

International

Seafood

Sustainability

Foundation,

Washington, D.C., USA. Bakun, A., 1996. Patterns in the ocean: ocean processes and marine population dynamics. California Sea Grant College System/NOAA/Centro de Investigaciones Biologicas del Noroeste, La Paz, Mexico. 323 pp., ISBN 1-888691-01-8. Bard, F., Hervé, A., 1994. Structure du stock de l’albacore (Thunnus albacares) atlantique d’après les marquages comparés aux lieux de ponte. Rec. Doc. Sci 42, 204208.

25

Bard, F.X., Cayré, P., Diouf, T., 1993. Migrations, Resources, fishing and biology of the tropical tunas of the Eastern Central Atlantic. Food and Agriculture Organization of the United Nations. Baske, A., Gibbon, J., Benn, J., Nickson, A., 2012. Estimating the use of drifting Fish Aggregation Devices (FADs) around the globe. PEW Environmental group, discussion paper, 8p. Bates, D.M., 2010. lme4: Mixed-effects modeling with R. Springer. Benitez-Nelson, C.R., Bidigare, R.R., Dickey, T.D., Landry, M.R., Leonard, C.L., Brown, S.L., Nencioli, F., Rii, Y.M., Maiti, K., Becker, J.W., Bibby, T.S., Black, W., Cai, W.-J., Carlson, C.A., Chen, F., Kuwahara, V.S., Mahaffey, C., McAndrew, P.M., Quay, P.D., Rappé, M.S., Selph, K.E., Simmons, M.P., Yang, E.J., 2007. Mesoscale Eddies Drive Increased Silica Export in the Subtropical Pacific Ocean. Science 316, 1017-1021. Bertrand, A., Bard, F.X., Josse, E., 2002a. Tuna food habits related to the micronekton distribution in French Polynesia. Marine Biology 140, 1023-1037. Bertrand, A., Josse, E., Bach, P., Gros, P., Dagorn, L., 2002b. Hydrological and trophic characteristics of tuna habitat: consequences on tuna distribution and longline catchability. Canadian Journal of Fisheries and Aquatic Sciences 59, 1002-1013. Bjornstad, O., 2009. ncf: spatial nonparametric covariance functions. R package version 1.1-3. See http://onb. Ent. Psu. Edu/onb1/r.

26

BjØrnstad, O.N., Falck, W., 2001. Nonparametric spatial covariance functions: Estimation and testing. Environmental and Ecological Statistics 8, 53-70. Brehmer, P., Josse, E., Nøttestad, L., 2011. Evidence that whales (Balaenoptera borealis) visit drifting fish aggregating devices: do their presence affect the processes underlying fish aggregation? Marine Ecology. Cardinale, M., Linder, M., Bartolino, V., Maiorano, L., Casini, M., 2009. Conservation value of historical data: reconstructing stock dynamics of turbot during the last century in the Kattegat-Skagerrak. Marine Ecology Progress Series 386, 197-206. Castro, J., Santiago, J., Santana-Ortega, A., 2002. A general theory on fish aggregation to floating objects: an alternative to the meeting point hypothesis. Reviews in Fish Biology and Fisheries 11, 255-277. Cochrane, K., De Young, C., Soto, D., Bahri, T., 2009. Climate change implications for fisheries and aquaculture. FAO Fisheries and Aquaculture Technical Paper 530, 212. Chassot, E., Bonhommeau, S., Reygondeau, G., Nieto, K., Polovina, J.J., Huret, M., Dulvy, N.K., Demarcq, H., 2011. Satellite remote sensing for an ecosystem approach to fisheries management. ICES Journal of Marine Science: Journal du Conseil 68, 651666. Dagorn, L., Filmalter, J.D., Forget, F., Amandè, M.J., Hall, M.A., Williams, P., Murua, H., Ariz, J., Chavance, P., Bez, N., 2012a. Targeting bigger schools can reduce ecosystem impacts of fisheries. Canadian Journal of Fisheries and Aquatic Sciences 69.

27

Dagorn, L., Holland, K., Hallier, J., Taquet, M., Moreno, G., Sancho, G., Itano, D., Aumeeruddy, R., Girard, C., Million, J., 2006. Deep diving behavior observed in yellowfin tuna (Thunnus albacares). Aquatic Living Resources 19, 85-88. Dagorn, L., Holland, K.N., Restrepo, V., Moreno, G., 2012b. Is it good or bad to fish with FADs? What are the real impacts of the use of drifting FADs on pelagic marine ecosystems? Fish and Fisheries, n/a-n/a. Delgado de Molina, A., Ariz, J., Santana, J.C., Pallares, P., 2005. Análisis de los datos de marcado de patudo en las Islas Canarias. Collect. Vol. Sci. Pap, ICCAT 57, 100-115. Delgado de Molina, A., Rojo, V., Fraile-Nuez, E., Ariz, J., 2014. Analysis of the Spanish tropical purse-seine fleet’s exploitation of a concentration of skipjack (Katsuwonus pelamis) in the Mauritania zone in 2012. Collect. Vol. Sci. Pap. ICCAT 70, 2771-2786. Dempster, T., Kingsford, M.J., 2003. Homing of pelagic fish to fish aggregation devices (FADs): the role of sensory cues. Marine Ecology Progress Series 258, 213-222. Doray, M., Josse, E., Gervain, P., Reynal, L., Chantrel, J., 2007. Joint use of echosounding, fishing and video techniques to assess the structure of fish aggregations around moored Fish Aggregating Devices in Martinique (Lesser Antilles). Aquatic Living Resources 20, 357-366. Doray, M., Petitgas, P., Nelson, L., Mahévas, S., Josse, E., Reynal, L., 2009. The influence of the environment on the variability of monthly tuna biomass around a moored, fish-aggregating device. ICES Journal of Marine Science: Journal du Conseil 66, 1410-1416. 28

Dormann, C.F., Elith, J., Bacher, S., Buchmann, C., Carl, G., Carré, G., Marquéz, J.R.G., Gruber, B., Lafourcade, B., Leitão, P.J., Münkemüller, T., McClean, C., Osborne, P.E., Reineking, B., Schröder, B., Skidmore, A.K., Zurell, D., Lautenbach, S., 2013. Collinearity: a review of methods to deal with it and a simulation study evaluating their performance. Ecography 36, 27-46. Dubroca, L., Chassot, E., Floch, L., Demarcq, H., Assan, C., Delgado de Molina, A., 2013. Seamounts and tuna fisheries: Tuna hotspots or fishermen habits? Collect. Vol. Sci. Pap. ICCAT 69, 2087-2102. Edwards, E., 1992. Energetics of associated tunas and dolphins in the eastern tropical Pacific Ocean: a basis for the bond. Fishery Bulletin 90, 678-690. Farrell, E.R., Boustany, A.M., Halpin, P.N., Hammond, D.L., 2014. Dolphinfish (Coryphaena hippurus) distribution in relation to biophysical ocean conditions in the northwest Atlantic. Fisheries Research 151, 177-190. Fiedler, P.C., Bernard, H.J., 1987. Tuna aggregation and feeding near fronts observed in satellite imagery. Continental shelf research 7, 871-881. Floch, L., Delgado de Molina, A., Assan, C., Dewals, P., Areso, J., Chassot, E., 2012. Statistics of the european purse seine fishing fleet and associated flags targeting tropical tunas in the Indian Ocean (1981-2011). IOTC Working Party on Tropical Tunas, 14th session. IOTC, Grand Baie, Mauritius, 24-29 October 2012. Fonteneau, A., 1991. Monts sous-marins et thons dans l'Atlantique tropical est. Aquatic Living Resources 4, 13-25.

29

Fonteneau, A., Chassot, E., Bodin, N., 2013. Global spatio-temporal patterns in tropical tuna purse seine fisheries on drifting fish aggregating devices (DFADs): Taking a historical perspective to inform current challenges. Aquatic Living Resources 26, 37-48. Fonteneau, A., Lucas, V., Tewkai, E., Delgado, A., Demarcq, H., 2008. Mesoscale exploitation of a major tuna concentration in the Indian Ocean. Aquatic Living Resources 21, 109-121. Fonteneau, A., Marcille, J., 1993. Resources, fishing and biology of the tropical tunas of the Eastern Central Atlantic. Food and Agriculture Organization of the United Nations. Fonteneau, A., Soubrier, P., 1996. Interactions between tuna fisheries: a global review with specific examples from the Atlantic ocean. Status of Interactions of Pacific Tunas Fisheries in 1995. In Proceedings of the Second FAO Expert Consultation on Interactions of Pacific Tuna Fisheries, 23 – 31 January, Shimizu, Japan. 612 pp. Forget, F.G., Capello, M., Filmalter, J.D., Govinden, R., Soria, M., Cowley, P.D., Dagorn, L., 2015. Behaviour and vulnerability of target and non-target species at drifting fish aggregating devices (FADs) in the tropical tuna purse seine fishery determined by acoustic telemetry. Canadian Journal of Fisheries and Aquatic Sciences, 1-8. Fraile, I., Murua, H., Goñi, N., Caballero, A., 2010. Effects of environmental factors on catch rates of FAD-associated yellowfin (Thunnus albacares) and skipjack (Katsuwonus pelamis) tunas in the western Indian Ocean. IOTC-2010-WPTT-46.

30

Ghazali, S.M., Montgomery, J.C., Jeffs, A.G., Ibrahim, Z., Radford, C.A., 2013. The diel variation and spatial extent of the underwater sound around a fish aggregation device (FAD). Fisheries Research. Giannoulaki, M., Iglesias, M., Tugores, M.P., Bonanno, A., Patti, B., De Felice, A., Leonori, I., Bigot, J.L., TiČIna, V., Pyrounaki, M.M., Tsagarakis, K., Machias, A., Somarakis, S., Schismenou, E., Quinci, E., Basilone, G., Cuttitta, A., Campanella, F., Miquel, J., OÑAte, D., Roos, D., Valavanis, V., 2013. Characterizing the potential habitat of European anchovy Engraulis encrasicolus in the Mediterranean Sea, at different life stages. Fisheries Oceanography 22, 69-89. Gilman, E., Chaloupka, M., Read, A., Dalzell, P., Holetschek, J., Curtice, C., 2012. Hawaii longline tuna fishery temporal trends in standardized catch rates and length distributions and effects on pelagic and seamount ecosystems. Aquatic Conservation: Marine and Freshwater Ecosystems 22, 446-488. Goñi, N., Ruiz, J., Murua, H., Santiago, J., Krug, I., Sotillo de Olano, B., Gonzalez de Zarate, A., Moreno, G., Murua, J., 2015. System of verification of the code of good practices on board ANABAC and OPAGAC tuna purse seiners and preliminary results for the Atlantic Ocean. IOTC–2015–WPEB11–INF09. Grafarend, E.W., 2006. Linear and nonlinear models: fixed effects, random effects, and mixed models. . Walter de Gruyter. Hallier, J., Gaertner, D., 2008. Drifting fish aggregation devices could act as an ecological trap for tropical tuna species. Marine Ecology Progress Series 353, 255-264.

31

Hassrick, J.L., Henderson, M.J., Huff, D.D., Sydeman, W.J., Sabal, M.C., Harding, J.A., Ammann, A.J., Crandall, E.D., Bjorkstedt, E.P., Garza, J.C., Hayes, S.A., 2016. Early ocean distribution of juvenile Chinook salmon in an upwelling ecosystem. Fisheries Oceanography 25, 133-146. Hastie, T.J., Tibshirani, R.J., 1990. Generalized additive models. Monographs on Statistics and Applied Probability 43. . Chapman & Hall/CRC, London. Hernández-García, V., Santana Ortega, A.T., Ganzedo, U., Castro, J., 2014. Analysis of impact of non-entangling FADs on incidental catches in the Indian Ocean tuna fishery. IOTC–2014–WPTT16–18. Holland, K.N., Grubbs, R.D., 2008. Fish Visitors to Seamounts: Tunas and Bill Fish at Seamounts, Seamounts: Ecology, Fisheries & Conservation. Blackwell Publishing Ltd, pp. 189-201. Jones, A.R., Hosegood, P., Wynn, R.B., De Boer, M.N., Butler-Cowdry, S., Embling, C.B., 2014. Fine-scale hydrodynamics influence the spatio-temporal distribution of harbour porpoises at a coastal hotspot. Progress in Oceanography 128, 30-48. Josse, E., Dagorn, L., Bertrand, A., 2000. Typology and behaviour of tuna aggregations around fish aggregating devices from acoustic surveys in French Polynesia. Aquatic Living Resources 13, 183-192. Kahru, M., Fiedler, P.C., Gille, S.T., Manzano, M., Mitchell, B.G., 2007. Sea level anomalies control phytoplankton biomass in the Costa Rica Dome area. Geophysical Research Letters 34, L22601.

32

Kakuma, S., 2000. Synthesis on moored FADs in the North West Pacific region. Pêche thonière et dispositifs de concentration de poissons, Caribbean-Martinique, 15-19 Oct 1999. Lan, K.-W., Lee, M.-A., Lu, H.-J., Shieh, W.-J., Lin, W.-K., Kao, S.-C., 2011. Ocean variations associated with fishing conditions for yellowfin tuna (Thunnus albacares) in the equatorial Atlantic Ocean. ICES Journal of Marine Science: Journal du Conseil 68, 1063-1071. Lan, K.W., Evans, K., Lee, M.A., 2012. Effects of climate variability on the distribution and fishing conditions of yellowfin tuna (Thunnus albacares) in the western Indian Ocean. Climatic Change, 1-15. Lo, N.C.-h., Jacobson, L.D., Squire, J.L., 1992. Indices of Relative Abundance from Fish Spotter Data based on Delta-Lognormal Models. Canadian Journal of Fisheries and Aquatic Sciences 49, 2515-2526. Longhurst, A., 2006. Ecological geography of the sea Second ed. Academic Press, Elsevier. ISBN 978-0-12-455521-1. Longhurst, A., Pauly, D., 1987. Ecology of tropical oceans. Elsevier, ISBN: 978-0-12455562-4. 407p. Lopez-Calderon, J., Manzo-Monroy, H., Santamaria-del-Angel, E., Castro, R., Gonzalez-Silvera, A., Millan-Nunez, R., 2006. Mesoscale variability of the Mexican Tropical Pacific using TOPEX and SeaWiFS data. Ciencias marinas 32.

33

Lopez, J., 2015. Behaviour of tuna and non-tuna species at fish aggregating devices (FADs), ascertained through fishers’ echo-sounder buoys: implications for conservation and management. Department of Zoology and Animal Cell Biology, University of the Basque Country. 151 p. Lopez, J., Moreno, G., Boyra, G., Dagorn, L., 2016. A model based on data from echosounder buoys to estimate biomass of fish species associated with fish aggregating devices. Fishery Bulletin 114, 166-178. Lopez, J., Moreno, G., Sancristobal, I., Murua, J., 2014. Evolution and current state of the technology of echo-sounder buoys used by Spanish tropical tuna purse seiners in the Atlantic, Indian and Pacific Oceans. Fisheries Research 155, 127-137. Lopez, J., Moreno, G., Soria, M., Cotel, P., Dagorn, L., 2010. Remote discrimination of By-catch in purse seine fishery using fishers' echo-sounder buoys. IOTC-2010-WPEB03. Louzao, M., Bécares, J., Rodríguez, B., Hyrenbach, K.D., Ruiz, A., Arcos, J.M., 2009. Combining vessel-based surveys and tracking data to identify key marine areas for seabirds. Marine Ecology Progress Series 391, 183-197. Maclennan, D., Fernandes, P., Dalen, J., 2002. A consistent approach to definitions and symbols in fisheries acoustics. ICES Journal of Marine Science 59, 365. Marsac, F., 2013. Outline of climate and oceanographic conditions in the Indian Ocean: an update to August 2013 IOTC–2013–WPTT15–09.

34

Marsac, F., Fonteneau, A., Ménard, F., 2001. Drifting FADs used in tuna fisheries: an ecological trap? ACTES DE COLLOQUES-IFREMER, 537-552. Matsumoto, T., Satoh, K., Semba, Y., Toyonaga, M., 2016. Comparison of the behavior of skipjack (Katsuwonus pelamis), yellowfin (Thunnus albacares) and bigeye (T. obesus) tuna associated with drifting FADs in the equatorial central Pacific Ocean. Fisheries Oceanography, n/a-n/a. Matsumoto, T., Satoh, K., Toyonaga, M., 2014. Behavior of skipjack tuna (Katsuwonus pelamis) associated with a drifting FAD monitored with ultrasonic transmitters in the equatorial central Pacific Ocean. Fisheries Research 157, 78-85. Maufroy, A., Kaplan, D.M., Bez, N., De Molina, A.D., Murua, H., Floch, L., Chassot, E., 2016. Massive increase in the use of drifting Fish Aggregating Devices (dFADs) by tropical tuna purse seine fisheries in the Atlantic and Indian oceans. ICES Journal of Marine Science: Journal du Conseil. Maunder, M.N., Sibert, J.R., Fonteneau, A., Hampton, J., Kleiber, P., Harley, S.J., 2006. Interpreting catch per unit effort data to assess the status of individual stocks and communities. ICES Journal of Marine Science: Journal du Conseil 63, 1373-1385. McGlade, J.M., Cury, P., Koranteng, K.A., Hardman-Mountford, N., 2002. The Gulf of Guinea Large Marine Ecosystem: environmental forcing and sustainable development of marine resources. Newnes. Ménard, F., Marsac, F., Bellier, E., Cazelles, B., 2007. Climatic oscillations and tuna catch rates in the Indian Ocean: a wavelet approach to time series analysis. Fisheries Oceanography 16, 95-104. 35

Morato, T., 2006. Ecology and fisheries of seamount ecosystems. The University of British Columbia. Moreno, G., Dagorn, L., Sancho, G., Itano, D., 2007a. Fish behaviour from fishers knowledge: the case study of tropical tuna around drifting fish aggregating devices (DFADs). Canadian Journal of Fisheries and Aquatic Sciences 64, 1517-1528. Moreno, G., Josse, E., Brehmer, P., Nøttestad, L., 2007b. Echotrace classification and spatial distribution of pelagic fish aggregations around drifting fish aggregating devices (DFAD). Aquatic Living Resources 20, 343-356. Morrow, R., Birol, F., Griffin, D., Sudre, J., 2004. Divergent pathways of cyclonic and anti-cyclonic ocean eddies. Geophysical Research Letters 31, L24311. Mugo, R.M., Saitoh, S.-I., Takahashi, F., Nihira, A., Kuroyama, T., 2014. Evaluating the role of fronts in habitat overlaps between cold and warm water species in the western North Pacific: A proof of concept. Deep Sea Research Part II: Topical Studies in Oceanography 107, 29-39. Murua, H., Eveson, J.P., Marsac, F., 2015. The Indian Ocean Tuna Tagging Programme: Building better science for more sustainability. Fisheries Research 163, 16. Naimi, B., 2015. usdm: Uncertainty analysis for species distribution models, R package version 1.1-12.

36

Ohta, I., Kakuma, S., 2005. Periodic behavior and residence time of yellowfin and bigeye tuna associated with fish aggregating devices around Okinawa Islands, as identified with automated listening stations. Marine Biology 146, 581-594. Potier, M., Bach, P., Ménard, F., Marsac, F., 2014. Influence of mesoscale features on micronekton and large pelagic fish communities in the Mozambique Channel. Deep Sea Research Part II: Topical Studies in Oceanography 100, 184-199. Prince, E.D., Luo, J., Phillip Goodyear, C., Hoolihan, J.P., Snodgrass, D., Orbesen, E.S., Serafy, J.E., Ortiz, M., Schirripa, M.J., 2010. Ocean scale hypoxia-based habitat compression of Atlantic istiophorid billfishes. Fisheries Oceanography 19, 448-462. Robert, M., Dagorn, L., Deneubourg, J.L., 2014. The aggregation of tuna around floating objects: What could be the underlying social mechanisms? Journal of Theoretical Biology 359, 161-170. Robert, M., Dagorn, L., Lopez, J., Moreno, G., Deneubourg, J.-L., 2013. Does social behavior influence the dynamics of aggregations formed by tropical tunas around floating objects? An experimental approach. Journal of Experimental Marine Biology and Ecology 440, 238-243. Roessig, J., Woodley, C., Cech, J., Jr., Hansen, L., 2004. Effects of global climate change on marine and estuarine fishes and fisheries. Reviews in Fish Biology and Fisheries 14, 251-275. Roger, C., 1994. Relationships among yellowfin and skipjack tuna, their prey-fish and plankton in the tropical western Indian Ocean. Fisheries Oceanography 3, 133-141.

37

Rouyer, T., Fromentin, J.-M., Ménard, F., Cazelles, B., Briand, K., Pianet, R., Planque, B., Stenseth, N.C., 2008. Complex interplays among population dynamics, environmental forcing, and exploitation in fisheries. Proceedings of the National Academy of Sciences 105, 5420-5425. Santiago, J., 1998. The North Atlantic Oscillation and recruitment of temperate tunas. ICCAT Col. Vol. Sci. Pap 48, 240-249. Scott, G.P., Lopez, J., 2014. The use of FADs in tuna fisheries. European Parliament. Policy Department B: Structural and Cohesion Policies: Fisheries IP/B/PECH/IC/2013123, 70p. Schaefer, K., Fuller, D., 2013. Simultaneous behavior of skipjack (Katsuwonus pelamis), bigeye (Thunnus obsesus), and yellowfin (T. albacares) tunas, within large multi-species aggregations associated with drifting fish aggregating devices (FADs) in the equatorial eastern Pacific Ocean. Marine Biology, 1-10. Simmonds, E.J., MacLennan, D.N., 2005. Fisheries Acoustics: Theory and Practice, 2nd edn. London: Blackwell Science.437 pp. Song, L., Zhou, J., Zhou, Y., Nishida, T., Jiang, W., Wang, J., 2009. Environmental preferences of bigeye tuna, Thunnus obesus, in the Indian Ocean: an application to a longline fishery. Environmental Biology of Fishes 85, 153-171. Song, L.M., Zhang, Y.U., Xu, L.X., Jiang, W.X., Wang, J.Q., 2008. Environmental preferences of longlining for yellowfin tuna (Thunnus albacares) in the tropical high seas of the Indian Ocean. Fisheries Oceanography 17, 239-253.

38

Stenseth, N.C., Mysterud, A., Ottersen, G., Hurrell, J.W., Chan, K.-S., Lima, M., 2002. Ecological Effects of Climate Fluctuations. Science 297, 1292-1296. Stramma, L., Johnson, G.C., Sprintall, J., Mohrholz, V., 2008. Expanding OxygenMinimum Zones in the Tropical Oceans. Science 320, 655-658. Stramma, L., Prince, E.D., Schmidtko, S., Luo, J., Hoolihan, J.P., Visbeck, M., Wallace, D.W.R., Brandt, P., Kortzinger, A., 2012. Expansion of oxygen minimum zones may reduce available habitat for tropical pelagic fishes. Nature Clim. Change 2, 33-37. Stretta, J.M., 1993. Environment and tuna fishing in the eastern tropical Atlantic, Resources, fishing and biology of the tropical tunas of the Eastern Central Atlantic. Food and Agriculture Organization of the United Nations. Sund, P.N., Blackburn, M., Williams, F., 1981. Tunas and their environment in the Pacific Ocean: a review. Oceanogr. Mar. Biol. Ann. Rev 19, 443-512. Taquet, M., Sancho, G., Dagorn, L., Gaertner, J., Itano, D., Aumeeruddy, R., Wendling, B., Peignon, C., 2007. Characterizing fish communities associated with drifting fish aggregating devices (FADs) in the Western Indian Ocean using underwater visual surveys. Aquatic Living Resources 20, 331-341. Tew Kai, E., Marsac, F., 2010. Influence of mesoscale eddies on spatial structuring of top predators’ communities in the Mozambique Channel. Progress in Oceanography 86, 214-223. Torres-Irineo, E., Gaertner, D., de Molina, A.D., Ariz, J., 2011. Effects of time-area closure on tropical tuna purse-seine fleet dynamics through some fishery indicators

39

doi:10.1051/alr/2011143. Aquatic Living Resources FirstView, null-null. Ushioda, M., 2015. Estimating The Use of FADS Around the World. PEW Environmental group, discussion paper, 19 p. Wilson, C., Adamec, D., 2001. Correlations between surface chlorophyll and sea surface height in the tropical Pacific during the 1997–1999 El Niño-Southern Oscillation event. Journal of Geophysical Research: Oceans 106, 31175-31188. Wood, S., 2006. Generalized additive models: an introduction with R. . Chapman & Hall/CRC. Wood, S., 2011. gamm4: Generalized additive mixed models using mgcv and lme4. R package version 0.1–2. Wood, S., 2014. Package ‘mgcv’. R package version1.7-29. Wood, S.N., 2008. Fast stable direct fitting and smoothness selection for generalized additive models. Journal of the Royal Statistical Society: Series B (Statistical Methodology) 70, 495-518. Wood, S.N., Augustin, N.H., 2002. GAMs with integrated model selection using penalized regression splines and applications to environmental modelling. Ecological Modelling 157, 157-177. Zagaglia, C.R., Lorenzzetti, J.A., Stech, J.L., 2004. Remote sensing data and longline catches of yellowfin tuna (Thunnus albacares) in the equatorial Atlantic. Remote sensing of environment 93, 267-281.

40

Zarauz, L., Irigoien, X., Fernandes, J.A., 2008. Modelling the influence of abiotic and biotic factors on plankton distribution in the Bay of Biscay, during three consecutive years (2004–06). Journal of Plankton Research 30, 857-872. Zuur, A., Ieno, E.N., Walker, N., Saveliev, A.A., Smith, G.M., 2009a. Mixed effects models and extensions in ecology with R. Springer, New York. Zuur, A.F., Mira, A., Carvalho, F., Ieno, E.N., Saveliev, A.A., Smith, G.M., Walker, N.J., 2009b. Negative Binomial GAM and GAMM to Analyse Amphibian Roadkills, Mixed effects models and extensions in ecology with R. Springer New York, New York, NY, pp. 383-397.

Table 1. Tuna and tuna-like environmental preferences from other scientific studies (review table). + = Positive correlation between variable and tuna abundance ; - = negative correlation. [SKJ: skipjack (Katsuwonus pelamis); YFT: yellowfin tuna (Thunnus albacares); BET: bigeye tuna (Thunnus obesus); TT: tropical tunas (SKJ, YFT, BET); LP: large pelagics; DF: dolphinfish (Coryphaena hyppurus); WeA: western equatorial Atlantic; AO: Atlantic Ocean; NWA: north western Atlantic; WIO: western Indian Ocean; MZC: Mozambique Channel; LL: longline catches; Var c.d.: various catch data; Exp: experimental; PS: purse seine catches (PS FAD: FAD catches, PS FS: free swimming school catches); T: sea temperature; Chl: chlorophyll; S: salinity; O2: oxygen; TD: thermocline depth; WG: geostrophic current; Mig: migrations; Rep: reproduction; IOD: Indian Ocean Dipole; LSCI: Large-scale Climate Indices; NAO: North Atlantic Oscillation; ITCZ: Inter-tropical Convergence Zone; SI: social interaction; GoG: Gulf of Guinea]

Speci es

Are a

Dat a

Andrade, 2004

SKJ

We A

LL

Bard 1993; Bard and Herve, 1994; Fonteneau and Soubrier, 1996

TT

AO

Var.

Brill 1994; Block 1997; Prince 2006, 2010; Stramma 2008, 2012

LP

AO

Exp.

YFT

PO

LL

Author

Dell, 2012

C T h l

S O . 2

Wi nd

Ed die s

T D

W G

M ig.

R ep .

LS CI

1

2

3

+ +

+

41

S I

Doray, 2009

TT

Cari bbe

Exp.

Farrel, 2014

DF

NW A

LL

Fonteneau, 2008

YFT

WI O

PS

+

Fraile, 2010

SKJ/ YFT

WI O

PS (FA D)

+

Lan, 2011

Adult YFT

We A

LL

+

Lan, 2012

Adult YFT

WI O

LL

-

Lopez, 2015

TT

WI O

Exp.

Marsac, 2013

SKJ

WI O

PS (FA D)

Maury, 2001

YFT

AO

LL / PS

Menard, 2007

BET/ YFT

WI O

LL / PS

Potier, 2014

Adult YFT

MZ C

LL

Robert, 2013; 2014

TT

WI O

Exp.

Santiago, 1998; Stenseth, 2002; Rouyer, 2008

TT

AO

Var. c.d.

Song, 2008

Adult YFT

IO

LL

-

Song, 2009

Adult BET

IO

LL

-

TT

WI O

Exp.

Stretta, 1993

YFT

AO

Var. c.d.

Tew Kai, 2010

Adult YFT

MZ C

PS (FS)

Zagaglia, 2004

Adult YFT

We A

LL

Stehfest, 2010

-

-

+

-

-

4 IO D

+

+

+

IO D +

-

+ N A O

5

+

+

+

6

+

+

+

-

7

IT C Z

42

1

Spawning all year in eq. area or NE South America in spring and summer 2

Juveniles westward

3

GoG recruitment area

4

Adults westward and eastward

5

NAO effect on recruitment

6

No significance difference between trajectories of FS and FAD tuna 7

Adults westward

Table 2. AIC for GAMM and GAM-equivalents, log-likelihood ratio tests for the DFAD-specific random effect variance, and percent deviance explained by the GAM-equivalents. d.f.: degrees of freedom, “LLR”: log-likelihood ratio.

AIC

Log-likelihood Ratio Test

% Deviance

GAMM

GAM

GAM(re)

LLR value

d.f.

p-value

Accounted for by GAM

GAM(re)

Non-tuna presence/absence

1975

2131

1888

285.1

1

<0.0001

14.9

28.7

Non-tuna biomass density

993

964

960

-4.66

1

1

2.35

5.37

Tuna presence/absence

2247

2409

2124

222.61

1

<0.0001

8.66

23.6

Tuna biomass density

6597

7019

6461

418.95

1

<0.0001

5.17

29.2

Table 3. Selected GAMM models for non-tuna and tuna species. “prs/abs”: logistic model for the presence/absence of biomass; “biomass”: model for observations with biomass density.[SST: sea surface temperature; T30: temperature at 30m depth; T100: temperature at 100m depth; Chl-a: chlorophyll; Chla18: chlorophyll 18 days before; D: depth (bathymetry); TD: thermocline depth; TG: thermocline gradient; WG: geostrophic current; WT: total current; SLA: sea level anomaly; -: not significant; x: not included in the model;]

Non-tunas

Tunas

Parameter

prs/abs

biomass

prs/abs

biomass

Adjusted r2

0.13

0.014

0.06

0.02

-

-

-

-

Deviance explained (%)

43

AIC score

1975

993

2247

6597

n

2618

566

2652

2226

Residual d.f.

2606

561

2636

2215

d.f.

p-value

d.f.

p-value

SST

2.554

< 0.001

1

< 0.05

T100

-

-

-

Chl-a

2.522

< 0.05

Chl-a18

-

D

d.f.

p-value

d.f.

p-value

2.258

< 0.01

-

-

-

-

-

1.456

< 0.05

-

-

1.937

< 0.05

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

1

< 0.05

TG

-

-

-

-

-

-

1

< 0.01

WG

-

-

-

-

1

< 0.01

-

-

WT

-

-

-

-

2.733

< 0.01

-

-

SLA

-

-

1.835

< 0.05

2.451

< 0.001

-

-

Speed

1.773

< 0.05

1

< 0.05

-

-

-

-

Bearing

1.558

< 0.01

-

-

1.356

< 0.05

1.895

< 0.001

Soaking time

2.382

< 0.01

-

-

2.827

< 0.001

2.332

< 0.001

x

x

x

x

-

-

2.262

< 0.05

Covariates

Non-tuna biomass

Figure 1. Spatial distribution of the samples taken by the echo-sounder buoys for the study period. Figure 2. Smoothed fits of covariates modelling the presence – absence of non-tuna species. The “rug” above the x-axis shows the distribution of the observed variable values. The y-axis represents the centred smooth term contribution to the model on the scale of the linear predictor. Dashed lines indicate approximate 95% confidence bounds. Figure 3. Smoothed fits of covariates modelling the biomass density of non-tuna species. The “rug” above the x-axis shows the distribution of the observed variable values. The y-axis represents the centred

44

smooth term contribution to the model on the scale of the linear predictor. Dashed lines indicate approximate 95% confidence bounds. Figure 4. Smoothed fits of covariates modelling the presence – absence of tunas. The “rug” above the xaxis shows the distribution of the observed variable values. The y-axis represents the centred smooth term contribution to the model on the scale of the linear predictor. Dashed lines indicate approximate 95% confidence bounds. Figure 5. Smoothed fits of covariates modelling the biomass density of tunas. The “rug” above the x-axis shows the distribution of the observed variable values. The y-axis represents the centred smooth term contribution to the model on the scale of the linear predictor. Dashed lines indicate approximate 95% confidence bounds.

Figure 6. A scheme of various enrichment zones in the tropical eastern Atlantic Ocean overlaid on a satellite image of VIIRS chlorophyll. Enrichment zones are principally due to the upwelling of deep water during boreal summer and localized along the coast and along the Equator. Credit for providing the image is given to: CLS (Collecte Localisation Satellite, France, https://www.cls.fr).

45

46

47

48

49