A monitoring programme using fisheries observers was implemented from August 2008 to June 2018 regarding small-scale longline fishing (a “mother ship” fleet) from the fishing ports of Manta (Ecuador). This programme represents a small percentage of the Ecuadorian fishing fleet (10% for 2022; ACUERDO Nro. MPCEIP-SRP-2021-0208-A https://vlex.ec/vid/mpceip-srp-2021-0208-876693938). The observer data were provided by PAT – Ec, which is part of the National Plan for the Fisheries Control of Ecuador (PNCP – Ec). Observers were trained in biological and fishing data collection by the undersecretary of Fishery Resources (SRP); using the “Programa Nacional de Observadores Pesqueros de Ecuador” (PROBECUADOR). The information recorded in 160 boats included: target fish, hook type, number and size, type of bait, number and species of turtle caught, coordinates and date (Manual para el observador a bordo, Ministerio de acuacultura y pesca). These 160 boats each made between 1 and 29 fishing trips per year (mean ± SE = 5.5 ± 5), and between 1 to 20 sets (the action and result of releasing the longline) per trip (mean 4.3 ± 3). Therefore, we analyzed 1,363 sets, representing 17,965 of fishing hours. The soaking time (number of hours of the set in the water) (between 5 and 45; mean ± SE= 13.2 ± 7) and the number of hooks per set (between 8 and 1,680; mean ± SE= 322 ± 263) were used to estimate the Catch Per Unit of Effort (CPUE) = (number of catches/hour/1,000 hooks).

Observers recorded the catch composition (target and bycatch species) per set. The fish species were categorized for the statistical analysis, such as: tunas (Thunus obesus, T. albacaren or T. alalunga), billfish (Xiphias gladius, Makaira nigricans or Istiophorus platypterus) and mahi-mahi (Coryphaena hippurus) (this clustering is similar to Parga et al., 2015). The hooks were grouped into two categories, circle (C14, C15 and C16) and J (J1, 2, 3, 4, 5, 7, 8, 34, 36, 38 and 40) (see Mituhasi and Hall, 2011 for more details). The baits were grouped into: fish, squid and small scombrids (Auxis spp.). The seasons of the year were grouped into: winter (from December to May) and summer (from June to November) ( Table 1 ). The setting depth depends on the season and varies between 12-16 meters in winter (mahi-mahi) and between 20 and 30 meters in summer (tuna and billfish). The characterization of the study longline fishing is detailed in Table S1 .

The total population of artisanal fishers is made up from 63,972 Ecuadorian fishers (Alava et al., 2015). Therefore, the minimum number of fishermen required to survey for the sample to be representative according to the population was 382 (Singh and Masuku, 2014). In total, 421 fishers participated in the survey: 51 answered the online version of the questionnaire and 370 the in-person questionnaire ( Table S2 ). Age and years of experience were also registered since previous studies have showed that the fishers’ perception may change according to these variables (Bender et al., 2014).

The questionnaire was divided into two parts (i) the fishers’ experience, type of fishing gear, bait and boat used, and sociodemographic variables ( Annex 1 ), and (ii) the fishers’ perceptions about the problem of sea turtle bycatch. In the first part, questions were included about: whether they had experienced the bycatch of sea turtles and its frequency; their perceptions of trends in sea turtle bycatch in the last ten years; what gear, types of baits, season, or hooks they think cause the most bycatch. In the second part of the questionnaire, we asked questions such us: Is such bycatch a nuisance? Do they have a major environmental impact? Or, are they not relevant?

The development of the questionnaire follows three differentiated steps that provide it with precision and reliability (Gracia et al., 2021):

After these validation processes (content and comprehension), the questionnaire was composed of 16 sociodemographic questions and 13 closed questions about bycatch. The average time taken to fill in the questionnaire was 15 minutes per fisher. The results of the content validation are shown in Annex 2 .

The sampling technique applied was convenience sampling (Emerson, 2015) as fishers were selected based on availability and willingness to take part. Fishers were previously informed about the purpose of the present study and research ethical principles were applied to protect their anonymity, dignity, rights and welfare throughout the whole research project. Collecting the information from the fishers took place from June 2020 to November 2020 in ports of the province of Manabí and Esmeraldas. Two strategies were used to recruit participants for the study. First, in-person surveys were conducted in the ports of P. Lopez, P. Cayo, Machalilla, Jaramijó, Salango, S. Lorenzo, Pampanal, Tambillo and Manta (see Annex 3 ). These ports include the main points of artisanal and industrial fishing catches in Manabí and Esmeraldas (Ecuador). The questionnaires were completed in the presence of the fishers (n=370), by the interviewer. Simultaneously, we conducted an online questionnaire during the same period using the Google Forms tool (n=51). The online questionnaire was also promoted by the fishers’ cooperatives and guilds of Manabí and Esmeraldas.

To account for any uncertainty and potential bias in this convenient sampling method, we took some precautions and applied controls before the data analyses (Maas et al., 2021), for example: (i) distributing the survey online and on paper; (ii) distributing the survey at 9 ports to achieve higher levels of diversity and sample size in our survey ( Annex 3 ); (iii) controlling demographic representativeness by comparing our survey sample to census data concerning the fishers, showing a high level of representativeness for age and a higher representation of artisanal groups compared to the overall fishing populations. In addition, sampling bias was avoided through; (iv) the participants’ self-identification with the type of fishing that the fishers practice, type of boat, and type of fishing gear used, (v) reliability analyses of scale (Cronbach’s Alpha for internal validity). Cronbach’s Alpha was above 0.70 for our scale, representing a good level of fit (e.g., Cortina, 1993).

With respect to observer logbook data, one GLM (observer model)was carried out where the Catch Per Unit of Effort (CPUE) was included as response variable. In the case of fishers’ perception one GLMM (fishers’ model) was performed where number of turtles bycatch by fishers was included as response variable (frequency of the fishers’ perception of bycatch). In the observer model, the target species (3 levels), hook type (2 levels, circle vs J-hook), bait type (3 levels), turtle species (5 levels), and season (2 levels) were included as factors. The year was included as a random factor (ten levels). In the fishers’ model, the target species (2 levels), fishing gear (3 levels), hook type (2 levels), bait type (3 levels), turtle species (4 levels), and time of year (2 levels) were included as factors. The port was included as a random factor (six levels). For the observer model, a normal distribution with an identity link function was used, while for the fisher’s survey model, Poisson distribution with a log link function was used. Fisher’s Least Significant Difference (LSD) post-hoc test was also applied to check for response differences among different levels of categorical variables. The most plausible models were selected by comparing Akaike’s information criterion (AIC) (Burnham and Anderson, 2002) following a backward procedure (Zuur et al., 2009). The statistical analyses were performed using InfoStats software.

There were more men (n = 417) than women (n = 4) among the respondents, as well as a predominance of the intermediate age-class (30-45). The participation of older people and artisanal fishers was lower for the online questionnaire than the in-person questionnaire. The specialized fishing gear most commonly used were the gillnet (51%), followed by the longline (44%) and trawl net (5%). Artisanal fishers represented 94% of the responders, while professional fishers (working for a company for a salary) represented only 6% of the questionnaire’s respondents.

According to the observers’ data, 1,838 sea turtles were captured incidentally during the study period ( Figure 1 ), of which 72.3% were olive ridley (Lepidochelys olivacea), 23.7% green (Chelonia mydas), 1.5% leatherback (Dermochelys coriacea), 1.3% hawksbill (Eretmochelys imbricata), 0.8% loggerhead (Caretta caretta) and 0.4% undetermined sea turtles. In contrast, fishers identified that the green turtle was the most frequently caught species (46%), whereas they stated the olive ridley and hawksbill turtles to be caught at a lower rate, 27% and 24.7% respectively ( Table S3 ). The mean number of turtles caught incidentally per set was 1.3 ± 1 (range between 1 and 13), with an average of 4.2 sea turtles caught incidentally per 1,000 hooks.

Regarding the factors that influence the incidental catch of sea turtles, according to the observers’ data (model 1), the results show the effect of the target species, the season, the type of bait and the turtle species ( Table 2 ). The results show significantly more bycatch in winter ( Figure 2A ) and when squid is used as bait ( Figure 2B ). As for the target species, results show significant differences between when the intended catch was mahi-mahi and when it was other species (tuna or billfish), with a higher bycatch percentage of sea turtles when mahi-mahi was the target species ( Figure 2C ). Finally, there are significant differences between the proportions of turtles captured, with a particular predominance of the olive ridley turtle ( Figure 2D ).

Regarding the factors that, according to the fishers, influence bycatch (model 2), the results show the effect of the type of hook, the season, the target species, the turtle species and the fishing gear ( Figure 3 ). The highest incidental catch rates occur with the J-hook ( Figure 3A ), in winter ( Figure 3B ) and on a longline ( Figure 3C ). The results also show that the highest rates of bycatch occur when the target species are pelagic fish ( Figure 3D ). Finally, there are significant differences between the proportions of turtles captured, with a particular predominance of the hawksbill turtle ( Figure 3E ).

The fishers’ perception of bycatch was heterogeneous throughout the participants sampled, highlighting that age and years of experience influence the perception of bycatch ( Annex 4 ). According to our results, 39% of the fishers answered that bycatch is an important environmental problem, while 32% and 29% answered that it is not relevant or is a nuisance, respectively. Their perception of the temporal trends in the last ten years of bycatch is that it remains constant, which is similar to data reported by the observers. Concerning the question of what fishing gear was associated with more accidental catches of turtles, 43% point to the longline; followed by the gillnet 31% and the trawl net 26%. With respect to the bait, squid was the bait most indicated by the observers as responsible for bycatch (60%), followed by fish (35%), which coincides with what is reported by fishers ( Table 3 ), where squid represented 57% and fish 40%. Both the data from the observers and the perception of the fishers show that winter is the season with the most bycatch. For the target species, 78% of fishers highlight that bycatch occurs when pelagic species (included mahi-mahi, tuna or billfish) are selected, while in the case of observers, mahi mahi was involved in 53% of the captures. As for the moment at which the bycatch occurs, 60% of the respondents stated that it occurs when the fishing gear is already set versus 14% hauling or 26% setting fishing gear ( Annex 4 ). In addition, 70% indicated that the bycatch was due to entanglement with fishing gear. Finally, 63% said the J-hook was responsible for the bycatch.

According to the observers’ reports, the main factors associated with turtle bycatch detected in this study were bait type, target species and season. However, the hook type was not found to be a significant factor, possibly because most J-hooks used in the EPO are slightly smaller than circle hooks, which possibly cause the same probability of being swallowed (Parga et al., 2015) but J-hooks cause much more damage and deaths (Parga, 2012). Previous studies such as Foster et al. (2012) in North Atlantic Ocean, Yokota et al. (2009) in Western North Pacific or Coelho et al. (2015) in Tropical Northeast Atlantic Ocean have already shown that squid increases bycatch as it is difficult to separate from the hook and it is swallowed whole (Stokes et al., 2011; Serafy et al., 2012). In the case of the target species, mahi-mahi is the main species linked to the capture of sea turtles, especially for the olive ridley turtle (Whoriskey et al., 2011, Costa Rica’s Pacific; Bugoni et al., 2008, Southern Brazil). The results also showed winter as the maximum catch period, which coincides with the peak of mahi-mahi catches as well as when fishing occurs (Whoriskey et al., 2011; Andraka et al., 2013). However, this may be related to the depth of the set, since the mahi-mahi fishing terminal gear is set at between 12-16m deep in winter, while in summer tuna and billfish are deeper, at 20-30m. Previous studies in Madeiran Waters (Dellinger and Ferreira, 2005), Mediterranean Sea (Alessandro and Antonello, 2010) or Atlantic and North Pacific (Swimmer et al., 2017) have shown that the main interaction depth with longlines is in the upper 20 m of the water column, especially when hooks are set between 10 and 15m deep, which coincides with the mahi-mahi relatively shallow fishing strategy.

Data from fishers were collected in 2020 and from fishery observers from 2008 to 2018. Data from observers and fishers coincide in that winter is the period of most bycatch and that pelagic species are the target catch when the bycatch occurs but differ in those fishers highlight the role of longlines as the most harmful fishing gear for turtles compared to trawl nets and gill nets ( Annex 4 ). This was also identified at other locations such as in the Equatorial Eastern Atlantic (Carranza et al., 2006) or US North Atlantic (Watson et al., 2005). With this gear, the turtle ingests the hook with the bait, up to 40% deep in the throat with the highest probability of mortality (Stokes et al., 2011; Parga, 2012; Stokes et al., 2012; Swimmer et al., 2014). According to the fishers’ perception, nets rank second in the causes of bycatch and the mortality of sea turtles, as in Pacific or Atlantic fisheries, especially at the highest levels by gillnets and trawl nets (Wallace et al., 2010; Wallace et al., 2013; Pingo et al., 2017; Awabdi et al., 2021). Regarding the type of hook, fishers do indicate that the J-hook is responsible for a higher bycatch rate compared to circle hook ( Annex 4 ), up to 19.3 turtles per 1,000 hooks for longlines (eastern Pacific Ocean) (Wallace et al., 2010), with the dimensions (length) of between 41 to 60 mm and 20 to 30 mm having the greatest rates of turtle bycatch (Caracappa et al., 2018).

The bycatch rates obtained in this study (4.3 turtles per 1,000 hooks) are comparable to those extrapolated from the results of Barragán et al. (2009) in Machalilla National Parks (Ecuador), which are approximately 4.8 turtles per 1,000 hooks, or those reported by Whoriskey et al. (2011) in Costa Rica’s Pacific, where bycatch rates were 9.05 per 1000 hooks for olive ridley turtle and 0.35 per 1000 hooks for green turtles. On the other hand, they are relatively higher than those reported by Wang et al. (2021) in the Pacific Ocean, Gilman et al. (2007) in Hawaii longline fleet or Jaiteh et al. (2021) in the western Pacific Ocean (Caroline Islands) with sea turtle bycatch rates ranging from 0 to 0.024/1,000 hooks, 0.094/1,000 hooks and 0.29/1,000 hooks, respectively.

In this study, the observer data indicated higher frequency of catch of green and olive ridley turtles, while the fishers’ data identified hawksbill turtles to be most frequently captured In this sense, it is important to train fishers in species identification to improve accuracy of fisheries data collection (Fulton et al., 2019). Awabdi et al. (2021) in southeastern Brazil also found that green turtle was the most-captured species with all fishing gear (gillnet, trawl net and longline) according to the fishers’ perception, and Huang and Liu (2010) highlighted the capture of the olive ridley turtle by tuna longline fleets in the Indian Ocean.

Quantitative inclusion of the human element can increase our understanding of marine conservation issues (Awabdi et al., 2021; Primack et al., 2021). Our results highlight the fact that age and years of experience modify the perception of bycatch, as Bender et al. (2014) observed. Additionally, other parameters such as educational or socioeconomic level also have an effect on perception (Pont et al., 2016). For instance, a more negative perception towards South American sea lions was found among less educated fishers who had no other source of income besides fishing. Sanguinetti et al. (2021) found that older fishers with less formal education have a high focus on maximizing fishing yield, while younger fishers have a more sustainable and conservationist view of fishing. Similarly, Liu et al. (2019), found that artisanal fishers’ perceptions of marine mammals were predicted by fishing experience and education level. Therefore, it is relevant to include these parameters in studies of this type.

This study has showed that only 39% of fishers perceive incidental catch as an environmental problem, 60.6% consider it as a nuisance or not relevant. This coincides with the results of Aguilar-González et al. (2014) in the Gulf of California, where individual fishers do not see themselves as part of the problem. Although data concerning perception could be biased, since fishers may consider bycatch to be a larger problem than they indicated in the questionnaires because of concerns regarding their livelihoods or negative previous experiences such as damaging nets, spoiling catch, removing bait, or the endangered status of the species (Godley et al., 1998). However, validating the fishers’ answers with data collected directly by observers on the fishing vessels could help to minimize this bias (Carreras et al., 2004). Finally, understanding how worldwide fishers perceive and use resources has significant management and policy implications (Awabdi et al., 2021), which would allow the development of robust management models for sustainable fishing, as pointed out by Karnad et al. (2014).

At the national level, the results showed the high-level impact that the artisanal fleet has on sea turtles, which was similar to previous studies (Alfaro-Shigueto et al., 2011; Alfaro-Shigueto et al., 2018); highlighting that the season (winter) and type of hooks used with mahi-mahi are associated with bycatch of sea turtles (Andraka et al., 2013). Although Ecuador is implementing measures to reduce the bycatch of sea turtles and other megafauna (such as adoption of circle hooks, distribution of tools and training to unhook turtles from hooks or fisheries observer programs (https://www.iattc.org/GetAttachment/), these policies are still insufficient or not implemented by fishers (Alava et al., 2019). For this reason, empowering fishers’ governance is crucial to mitigate megafauna bycatch. At an international level, the results coincide with other studies in the EPO region, which highlight that in order to expand the use of circle hooks in the EPO region, governments should guarantee the availability of circle hooks at competitive prices in each country, and fishery authorities should implement regulatory measures in the use of tools to handle and release turtles (Parga et al., 2015). Therefore, the certification of the implementation of sustainable practices should be presented to fishers as a chance to access international, environmentally sensitive markets. Likely, measures at the international level should invest on scientific data collecting, training, support for the role of women, access to new technologies (by-catch reduction technologies) and promote more sustainable fishing practices (similar to https://oceans-and-fisheries.ec.europa.eu/fisheries/rules/small-scale-fisheries_en).

Despite the large amount of information generated by the “Ecuador National Fisheries Observer Program”, it only represents a small percentage of the Ecuadorian fishing fleet (10% for 2022; ACUERDO Nro. MPCEIP-SRP-2021-0208-A). Onboard observers can only be efficiently placed on one of the two decks, and information is provided by the crew to complement it, underestimating bycatch estimates (Luck et al., 2020), especially when bycatch estimation is of a lower priority (Forget et al., 2021). The longline fleet is the only one monitored by fisheries observers, while other fishing gears (such as gillnets or trawls) are absent. Another limitation is the possible bias of the fishers when answering the questions due to the fact that their perceptions depends on their memory and bycatch numbers could not be accurate (Mustika et al., 2021). To avoid this, it is relevant to include educational or socioeconomic variables in this type of studies (Pont et al., 2016); and to carry out studies with holistic approaches, where research integrates insights from local fisher communities with large-scale, long-term monitoring programs (Vásquez-Carrillo and Peláez-Ossa, 2021). Natural sciences are increasingly recognizing the value of social science methods for conducting conservation research, through interdisciplinary collaboration (Lowe et al., 2009). In this sense, there is also a growing foray by ecologists into social science realms (Lowe et al., 2009; Moore et al., 2010; Awabdi et al., 2021). For this study, we found several obstacles to achieve our research objectives, which were solved by collaboration with social scientists.