Study area and context

Sainte-Marie Island harbours 17 town (16 of which located directly on the shorelines) arranged in four boroughs (Fig 1). The towns differ in site-level characteristics that might influence perceptions related to fisheries and individual’s ability to cope with changes. Notably, infrastructures (e.g., large hotels, airport, port, seafood processing factory) are concentrated in the main town, Ambodifotatra (latitude, longitude = -16.998462°, 49.852690°), and the southern boroughs, providing more access to alternate livelihood options. While all boroughs experience a climate favorable to cash crops and trees (i.e. clove trees, vanilla), their exposure to seasonal extreme weather events (heavy rainfalls, annual cyclones) and bad weather varies. In particular, the western boroughs, separated from Madagascar’s mainland by the 7-23km wide Sainte-Marie canal, are more sheltered than the eastern boroughs. The southeastern shoreline harbours most of the near-shore coral reefs and lagoons, which attract tourists and influence local fishing practices.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Map of Sainte-Marie Island, Madagascar.

The boundaries of the 17 towns (contour line), the four boroughs (shades of grey), as well as coastal lagoons (dotted), are shown. Black dots indicate the town sampled as follow; 1 = Agniribe, 2 = Vohilava, 3 = Mahavelo, 4 = Ambodiforaha (borough = Vohilava); 5 = St Joseph, 6 = Ankobaoba, 7 = Ambodifotatra, 8 = Agnalaradjy (borough = Ambodifotatra); 9 = Maromandia, 10 = Agnafiafy, 11 = Loukintsy, 12 = Sahasifotatra, 13 = Agnivorano (borough = Loukintsy); 14 = Ifotatra, 15 = Ambatourao, 16 = Ambodiatafana (borough = Ambatourao). Data layers: Coral reefs [42], Madagascar subnational administrative boundaries [43].

https://doi.org/10.1371/journal.pone.0259792.g001

Fishing practices and targets are extremely diversified. Most fishermen use non-motorized pirogues, and the main fishing techniques are line-fishing, netting (gill nets, beach seine), free diving, and fish traps. Fishers typically use several fishing techniques: technique use is influenced by fishers’ conditions, preferences, target species, the weather, and fishing locations. For instance, older fishers are more likely to use fish traps while many younger fishermen dive; under bad weather conditions, fishers are more likely to be found netting in the lagoons that line fishing offshore. Fishers generally harvest multiple species. They used more than 50 common names referring to the most commonly caught fish and marine resources (see S1 Table for the species caught the most frequently), which included both mobile (e.g. migratory fishes) and sedentary species (e.g. mollusks, arthropods). As for market, fishers mainly sell their catches to a commercial fishery with an international export market, hotels, restaurants, and locals. In most towns, local collectors working for the exporting fishery are given fridges to store marine resources, which are then collected, processed, and packaged for export. As many other small-scale fisheries, fishery participation is gendered [44–46]; fisherwomen tend to fish for subsistence, target fish species of lower economic value as well as invertebrates, and search for marine resources in easily accessible near-shore areas.

Sainte-Marie’s nearshore fishery is managed under national fishery laws, which are not fully complied with and enforced as responsible authorities lack the capacity to enforce them. It is worth noting that a protected area was established in 2014 by the Indian Ocean Commission and local stakeholders. It generated significant hostility from the fishers it impacted in the Southern borough. Eventually, the resistance of the communities led to conflicts and the resulting cancellation of the initiative. Other than western approaches to management, local cultural taboos (fady) may also play a role in regulating fishery-related activities. Some fady entail species-specific protection: killing Guitarfish is taboo for all, and eating sea turtles for certain families. Other taboos are location specific. For instance, on the sandy islets off the southeastern end of the island, harvesting several types of marine resources at once, fishing with nets, and polluting the area, are prohibited. Access is also restricted (prohibited to Merina, a highlander ethnic group). When taboos are violated, spiritual sanctions involve mischance, disease, or death; material sanction can also be ordered by the community. However, fishers express that taboo enforcement and respect has declined because of several external changes, including the apparition of a new market for guitarfish fins, market multiplication, and harder times. In conclusion, marine resources are essentially open access.

In 2015, the non-governmental organization GRET -Professionels du développement solidaire (professionals for fair development)- started working with communities to improve livelihood sustainability. This led to the creation of the PCADDISM -Plateforme de concertation des acteurs pour le développement durable de l’île Sainte-Marie (Concertation platform for sustainable development actors on Sainte Marie Island)-, a platform managed and led by communities where local stakeholders (e.g. local fishers’ federation, commercial fishery, private hotels) are represented. The GRET provides technical and financial support to the PCADISM. When the present research was conducted, the idea to create LMMAs governed by traditional laws (dina) had started to gain momentum within the PCADISM. The PCADISM was holding meetings in each town to discuss livelihood sustainability, including possible fishery management strategies. Those meetings were open to all who wished to come, highly deliberative, and widely attended.

Survey strategy

Between June and October 2017, we interviewed fishers in 16 of the 17 towns (Fig 1). We could not find any fishers in Ambohitry, the only town not located directly on the shorelines. As we found that fishers’ availability was weather-dependent (they usually were at sea when the weather was good), we conducted data collection during the raining season. Each town was visited at least three times to sample as many fishers as possible. Upon arrival in a town, we introduced ourselves to local authorities and explained the purpose and intended use of the survey. After obtaining their permission to conduct research in the area, we asked them to suggest areas where fishers were likely to be located and to identify suitable survey respondents. We asked interviewees to refer us to additional fishers (snowball sampling). Sampling was purposive rather than random as we wanted to ensure sufficient representation of various fishing practices and socioeconomic backgrounds. As part of this effort, we employed a gender-inclusive approach [45]. For instance, when looking for suitable respondents, we defined fishers broadly as people who extract marine resources using various fishing methods for commercial or subsistence purposes. We also stated that we were looking for both women and men. The interviews were semi-structured and conducted in Malagasy. After being told about the purpose of the survey and its intended use, interviewees were asked whether they consented to participate. Interviews lasted between 30min and 3 hours.

Interview structure

Our initial questions (S2 Table) focused on fishers’ socioeconomic characteristics and fishing history, for example, when they began fishing, what species they targeted, whether they used fishing vessels, and with what gear. Our next questions focused on fisher perceptions of changes, including whether they had noticed changes in fish abundance, fish sizes, or knew of local species extinction. We also asked whether they had altered their fishing effort and fishing locations.

The next part of the interview focused upon temporal trends in catch sizes and fishing ground location. Both empirical [47, 48] and cognitive studies [49, 50] showed that people typically have trouble recalling events and their timing unless provided with appropriate cues. Hence, we asked fishers to recall catch sizes for their primary target species (for up to three targets) during the past year [51]; we then asked fishers to recall their catches for the period when they first began, and around 2009 when applicable. We used 2009 as a cue because that year an intense cyclone caused tremendous damages on the island, thereby impacting the memory of its inhabitants. We focused on great catch sizes as we found that fishers had trouble estimating and recalling average catch sizes; furthermore, studies have shown that people tend to recall great or poor catch sizes more accurately that typical catch sizes [52]. We also asked fishers about their fishing site location (distance from the shore) during the past year, and for the period when they first began fishing.

We then asked fishers what, in their opinion, could be the underlying causes of the changes and what could be done to improve ecological health in the area. Finally, we determined fishers’ coping mechanisms to future changes by asking what they would do in response to a hypothetical scenario of further decline.

Ethical considerations

Data collection aimed to inform the type of interventions that a non-governmental association, the GRET, could deliver to help reduce poverties and inequalities locally. The GRET is service-oriented; it does not have an institutional ethics committee, and there is no human research ethics committee overseeing the activities of non-governmental associations in Madagascar. However, we followed published recommendations for ethical research conduct in Madagascar [53]. All the activities carried out by the GRET at Sainte-Marie (including this study) were reviewed and approved by the decentralized technical services of the State and local authorities prior to being conducted. We shared the results of our study with local stakeholders and at higher levels, and adhered to common principles of human research ethics [54, 55]. For instance, we obtained verbal consent from the research participants prior to conducting the survey, ensured that the surveyed population mirrored the population targeted by the initiative, and obtained verbal agreement from the community leaders to conduct research. The data presented in this study is anonymous and aggregated to prevent direct linkage with individuals.

Statistical analysis

We used generalized additive models (GAMs) to model temporal changes in best catch sizes for 13 of the most frequent target species family (local Malagasy name) identified from the interviews (S1 Table). GAMs were ideally suited to the structure of our data and the nature of our analysis because; 1) their non-parametric smoothing function (hereafter referred to as smoothers) allowed us to model nonlinear temporal trends [56, 57]; 2) they can incorporate both continuous and categorical variables; 3) they can accommodate random effects; and, 4) they estimate the shape of the relationship from the data itself (we did not have to specify any a-priori shape). For these reasons, GAM represented a flexible and powerful approach to model temporal trends in best catches, as well as their nature and timing. We modeled temporal trends with negative binomial GAMs and a logit link using the mgcv package [56] (S1 File for R script). A first model (Mod1) included different intercept for each species, a smoother for time, and its interaction with species. The species term corresponded to the common Malagasy name, which often applied to multiple fish species (S1 Table). As perceptions varied among fishermen, we fitted fisherman identity as a random intercept and slope. To investigate plausible alternative hypotheses, we constructed several additional candidate models. Mod2 included the species intercept and a smoother for time; Mod3 and Mod4 included either one of the species intercept or time smoother. We selected the best model based on the lowest AIC criterion, highest restricted log-likelihood, and highest explanatory power. Furthermore, we evaluated the significance of the fixed effect using Wald’s test.

To model temporal changes in fishing distance from the shore, we used generalized linear models with a Poisson distribution because fishing distances were positively skewed (S1 File for R script). As the magnitude of the perceived changes could vary by gear, we fitted a first model (Mod1b) with an interaction between fishing gear and time as a fixed effect. Mod2b included an intercept for gear and time; Mod 3 and 4 included either one of the gear intercept or time. We selected the best model based on the lower AIC and evaluated the significance of the fixed effect based on likelihood ratio-tests.

We then used Multiple Factor Analysis (MFA) to explore how an individual’s propensity to suggest management restrictions was associated with individual characteristics, site-level characteristics, coping mechanisms, perceptions related to ecological health and underlying causes. Tailored to accommodate qualitative variables, MFA is an extension of Principal Component Analysis in which variables within the same groups are weighted to balance the importance of the groups. It enabled us to account for the existing structure of our data, where we had several groups of variables; for instance, the group “perceptions of ecological health” included responses about local extinction, changes in size, fish numbers, and fishing site distance (Table 1). We performed the MFA using the “FactoMineR” package [58] in R [59] (S2 File for R script). We used Horn’s parallel analysis implemented in the “paran” package [60] to determine the number of dimensions retained in the MFA. We included support for restrictions as a supplementary variable: individuals that are close to one another in multidimensional space share similar individual attributes, perceptions of ecological health, underlying causes, coping mechanisms, or local assets. The main dimensions of this variability are then related to support for restrictions, specified as a supplementary variable. We used the v.test statistic to identify significant association (v.test >2) between support for restrictions and the dimensions [61]. To understand dimensional grouping and associations between variables, we used the visual representation of the MFA and identified the variables that were the most significantly associated (R²>0.20, p<0.05) with each dimension. Prior to running the MFA, we ran the Keyser-Meyer-Olkin (KMO) and Bartlett’s test of homogeneity of variances implemented in “psych” [62] to confirm sampling adequacy.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Groups of variables tested in the multiple factor analysis (MFA) and associated accronyms.

https://doi.org/10.1371/journal.pone.0259792.t001

Prior to running the MFA, answers had to be classified (Table 1). Questions related to perceptions of ecological health, underlying causes, and solutions were classified as positive and labelled as such when they could support management. For instance, when asking about solutions to the observed decline, fishers who suggested management restrictions as a mean to improve ecological health were classified as having a conservation-oriented attitude and labelled as “yes”, and those who did not suggest management restrictions were labelled as “no” (Table 1). Similarly, fishers who identified local fishing as a cause of the decline were labelled as “yes”, and those who did not as a “no”. For the coping mechanisms, we classified the responses to changes (past and future) as; 1) adapt in a way that could amplify a decline (i.e. fish harder, move location, change gear), 2) continue fishing as before, or 3) decrease fishing [63]. In Sainte-Marie, “adapt” is not a conservation-prone behaviour as it could amplify the scale of the decline by sequentially depleting marine resources [63, 64]. Similarly, “continue” could also amplify the decline; however, it is more of a coping mechanism than an adaptation. The site-level variables were codified based on whether they were under or above Sainte-Marie’s median value for each town using data from a socioeconomic survey of Sainte-Marie (GRET, unpublished data 2021).

Fishers’ characteristics

In total, we interviewed 127 fishers, including 109 men and 18 women (S3 Table). Despite our sampling efforts to collect additional data on women (including additional trips to specifically seek out fisherwomen), surveyed men outnumbered women. This result partly reflects the male dominance in the fishery. Indeed, based on a recent socioeconomic survey, we estimate our sample sizes to represent 2 and 4% of, respectively, Sainte-Marie’s fishermen and women (GRET, unpublished data, 2021).

As typical of many small-scale fisheries [45], fishing practices differed among surveyed fisherwomen and men (Table 2). Most men used more than one fishing gear; the predominant gear was line fishing (80%), followed by nets (50%), diving (28%), traps (16%), and spears (2%). Most women used nets (88%) followed by spears (12%). Men and women differed in fishing grounds; most men fished offshore from non-motorized pirogues (92%) while women generally fished on foot close to their towns (94%). These contrasting fishery practices are representative of gender-based differences practices in the area (GRET, unpublished data, 2021). The interview showed that men typically sold their catches to more than one market, of which the towns and the exporting fishery were the most frequently cited (82 and 60%, respectively); by contrast, women predominantly fished to feed their families. During data collection, we only encountered three migrant fishermen (two of which had been in Sainte-Marie for more than a decade).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Fishing practices and socioeconomic characteristics by gender for Sainte-Marie’s fishers.

https://doi.org/10.1371/journal.pone.0259792.t002

Surveyed men and women differed in their level of formal education, reported income, and investment in the fisheries. The number of women with no formal education was proportionally greater (60 vs 72%, respectively men and women). Women, who reported lower average incomes (122 VS 61 USD$), were also less likely to have alternative income sources (51 and 22% of, respectively, men and women). Men were more committed to the fishery; they were more frequently members of local fisher’s associations than women (64 vs. 6%) and owned expensive fishing gear (i.e boats, nets) more regularly (83 vs 39%). As differences in the circumstances and involvement of women emerging from this study and a socioeconomic survey of the area (GRET, unpublished data, 2021) could influence their perceptions, we discuss results separately among gender in the following section.

Temporal changes in catch sizes and fishing site distance

We found evidence for markedly nonlinear trends in best catch sizes that cannot be interpreted easily without fishers’ explanations of changes in the broader socio-ecological system and their own fishing practices, which are summarized in the discussion (Fig 2). Mod2 outperformed the other models: it yielded the lowest AIC, the higher log-restricted likelihood, and the variance explained (R² and deviance explained) was nearly as high as Mod1 (S4 Table). The F tests further confirmed this conclusion: the interaction between time and species family was insignificant (Wald test p-value = 0.58), while the time smoother and species family intercepts were significant (Wald test p-values < 0.01). Generally, the smooth term for time showed an increase in best catch sizes from 1964 to the mid-1990s (Fig 2). The increase was most pronounced between 1980 and 1990; from the mid-1990s to 2017, best catch sizes decreased. This reconstruction of best catch sizes was based on 562 observations, from 101 fishers (mean = 5.6 observations per fisherman), for the 13 most frequently caught Malagasy term for fish or marine resources (mean = 43.23 observations per species group).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Sainte-Marie’s fishermen perceptions of catch changes over time estimated from generalized additive model (GAM).

The solid red line is the fitted smooth function for time, the blue dotted lines represent the 95% confidence intervals, the rugs on the margins display the location of individual observations, and the residual points are shown. The y limits of the plot are set (-1<y<1.5) to facilitate the visual interpretation of trends in best catch sizes.

https://doi.org/10.1371/journal.pone.0259792.g002

The assessment of fishing site locations indicates a progressive change from nearshore to offshore fishing for some of the fishing gear (Fig 3). Mod1b outperformed the other models based on AIC and variance explained (S5 Table). The interaction between fishing gear and time was significant (deviance = -6095, df = 199, p<0.001). Notably, we found that the magnitude of the changes perceived were greater for those fishing with hooks, followed by nets, and then free divers. Alternatively, fishers using fish traps were now fishing closer to the shore.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Sainte-Marie’s fishermen reported changes in fishing distance from the shore.

Regression slopes indicate differences the trajectories of changes in fishing distance for each of the four main fishing gear (hook, net, diving, and traps) inferred from generalized linear models. The y limits of the plot are set (0<y<15,000) to facilitate the visual interpretation of trends in fishing distance from the shore.

https://doi.org/10.1371/journal.pone.0259792.g003

Fishers’ perceptions

Causes.

When fishers were explicitly asked about the causes of the changes (Table 3), they predominantly attributed it to local fishing (84 vs. 59% of men and women). Specifically, fishers mentioned an increased number of fishers competing for marine resources (60 and 33%) followed by the evolution of fishing techniques. In particular, many felt that net-fishing practices had become less selective and more destructive (34% of men). Beside local fishing, several fishers (24 and 11%) attributed the decline to overlapping fisheries: they referred to the commercial exploitation of shrimp in the canal (8% of men) and sea-cucumbers on the nearshore areas (20 and 11%). They described the sea cucumber fishers as migrants who sporadically visit Sainte-Marie. There was a strong perception that those fishers competed with local fishers by using diving cylinders to collect marine resources and did not contribute to the local economy. Finally, fishers mentioned several more general causes to the changes: in particular, many mentioned the socioeconomic circumstances (i.e. lack of other alternatives) (28% and 6% of men and women) and spiritual causes (6% and 44%).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Observed changes to the fisheries, inferred causes, suggested solutions, and coping mechanisms by gender for Sainte-Marie’s fishers.

https://doi.org/10.1371/journal.pone.0259792.t003

Relationship between fishers’ perceptions, individual- and site-level factors

Overall, the KMO analysis indicated sampling adequacy (S7 Table). The five dimensions recommended for retention under parallel analysis explained 53.48% of the total variance in the data, of which 28.71% was accounted for by the first two axes (S1 Fig). Support for restriction was significantly associated with dimension one (|v.test | = 4.82): however, it was not significantly associated with the remaining dimensions (|v.test statistic| = 1.43, 0.83, 0.86, 0.69).

Dimension one was mostly correlated with coping mechanisms (26.35%), individual characteristics (26.30%), and perceptions related to ecological health (25.32%) (Fig 4A, S8 Table). Variables most positively associated with the first dimension included variables related to social status (gender = women, no membership to fishing associations), individuals who were likely to have reduced their fishing efforts to cope with changes and responded that they would continue fishing in a hypothetical scenario of further decline, negative response to changes in ecological health (no local disappearance, changes in fishing site distances, or fish sizes) and underlying causal mechanism (not local fishing) (Fig 4A, S9 Table). Support for restriction was negatively associated with this dimension: it was associated with men, perceptions of impaired ecological health attributed to local human agency, and fishers who chose adaptation as a response to current and future changes in the fisheries. While individuals who supported fishery restrictions were clustered on the negative side of the first dimension, the coordinates of those who did not support fishery restrictions were spread across the MFA space (Fig 4B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Visualization of quantitative variables (a) and individuals (b) from a Multiple Factor Analysis (MFA) describing the relationship between support for conservation, individual, and contextal variables for dimension one and two.

On Fig 4A, Individual attributes (IN) include age (young/old), gender (woman/man), association membership (asso/noasso), attachment to fishing (att/noatt), dependence on fisheries (dep/nodep); perceptions related to ecological degradation (ED) include distance to fishing sites (dist/nodist), fish size (size/nosize), fish numbers (num/nonum), and local disapearance (disp/nodisp); perceptions related to underling causes (Co) associated with local fishing (fish/nofish); local assets (LA) include shops (sho/no), hotels (hot/nohot), leisures (leis/noleis), and lagoons (lag/nolag); coping mechanisms include realized (CR) to observed changes (continue/decrease/adapt) and hypothetical (CH) to further changes (continue/decrease/adapt). Support for restrictions (RE) is shown in red (yes/no). On Fig 4B, the small circles are individuals in support (blue) or not in support (red) for fishery restrictions; large symbols indicate mean group values and the elipses show the confidence intervals around group mean points.

https://doi.org/10.1371/journal.pone.0259792.g004

The second dimension was most represented by local assets (32.63%) and individual variables (29.49%). Variables positively associated with this dimension included younger individuals, local assets (lagoons, shops, leisure), continued fishing as a reaction to past changes (S10 Table). The third dimension was associated with perceptions related to ecological health (44.90%) and individual characteristics (26.66%). Variables most positively associated with this dimension included younger fishers and the perception that the number of fish had not declined (S11 Table). Dimension four was associated with coping mechanisms (58.20%) and individual characteristics (27.58%); specifically, individuals that responded that they would adapt to further changes in the fishery and those who were dependent on the fisheries were positively associated with this dimension (S12 Table). Alternatively, individuals that did not exclusively depend on the fisheries were associated with decreasing fishing effort as a response to further decline. Finally, dimension five was associated with local assets (42.09%) and individual characteristics (36.24%); specifically, fishers of towns with more hotels were less attached to fishing (S13 Table).

Long-term socio-ecological dynamics in Sainte-Marie

Our analysis of fishers’ perceptions highlights socio-ecological changes as a major force shaping the fishery’s temporal trajectory. Indeed, fishers described that the apparition of new markets, combined with limited livelihood options, drastically changed how locals interacted with the fishery. Specifically, more people began to fish, invest in fishing gear, and sell their catches. Due to these changes, fishery landings increased from the 1960s to the 1990s; however, this upward trend shifted in the 1990s. For instance, fishers began to perceive a decrease in fish abundance, catch sizes, and fish sizes; they witnessed the local disappearance of Dugongs (Dugong dugong), whose last official sighting in the area dates from 1993 [70], and two Mullidae who used to gather seasonally in large aggregations on Sainte-Marie’s shorelines. To sustain their catches, most fishers adapted their fishing practices; they fished longer, harder, with more efficient gear. Owing to the perception of reduced fish abundance in nearshore areas, many began to fish further offshore. Collectively, the socio-ecological changes described here can lead to the serial depletion of nearshore areas [64] and are known to drive a “race for fish” thought to be an important contributor to overfishing [21, 22, 71].

The consequences of this “race for fish” were indeed perceived by the fishers. For example, they identified an increased number of fishers, followed by the evolution of fishing gear (e.g. less selective, longer nets) as the main cause of the decline. Beyond local fishing, Sainte-Marie’s fishers also attributed the decline to overlapping fisheries. In particular, there was a strong perception that sea cucumber fishers, described as sporadic migrants using SCUBA-gear and motorized boats, compete with traditional fishers for marine resources. Although the use of SCUBA to collect marine resources is illegal in Madagascar, similar migrants operate without consequences in many areas [72]. Overall, the socioecological changes described here, as well as their timing, are in agreement with other studies of fishers’ perceptions and catch data in Madagascar [21, 22, 39, 71, 73–76]; collectively, these results suggest that socio-ecological changes occurring at both local- and national-scale contribute to the degradation of the fishery by increasing competition for marine resources.

Relationship between support for management restrictions, individual- and site-level factors

Overall, support for management restrictions was associated with variables implying individual commitment to the fisheries. For example, instead of reducing their fishing effort to develop alternative livelihoods, many fishers who suggested management restrictions had opted to adapt their fishing practices as a response to past changes; they tended to be members of local fishing associations, and shared an understanding of fishery changes and their underlying human causes. Those committed fishers -which were typically men- likely interpreted the current state of the fisheries as a threat to their livelihoods, thus providing an impetus to improve ecological health via fishery restrictions [9, 77, 78]. As such, our results align with other studies demonstrating a relationship between levels of personal bounding to fishery resources, awareness of ecological impairments, and conservation-oriented attitudes [12, 79, 80].

These conservation-oriented attitudes, however, were not associated with conservation-oriented coping mechanisms. Instead, they were associated with fishers who would respond to future changes by further adapting their fishing. This result, while seemingly contradictory, is consistent with studies showing that committed fishers may not be willing to exit fisheries [8, 81, 82]. This suggests that fishers who have adapted their fishing to past changes might displace their fishing effort as a response to future fishery restrictions, thereby coping with changes in a potentially unsustainable way even though they support fishery management restrictions [83].

Similarly, younger fishers were not associated with conservation-oriented coping mechanisms. In contrast to older fishers who were associated with decreased fishing effort, younger fishers continued fishing as a response to past changes. This result is contradictory to other studies where younger fishers were more adaptable and willing to exit the fisheries than their older counterparts [6, 84]. In Sainte-Marie’s context, continued fishing might be more of a coping strategy than an adaptive mechanism; younger fishers might therefore opt to reduce fishing if livelihood alternatives become available. Alternatively, younger fishers’ willingness to decrease fishing efforts might be influenced by shifting baselines [85]. Indeed, fishers generally believed that fisheries health was not as it once was; however, fishers who perceived no decline in fish abundance tended to be younger. Additionally, older fishers had experienced higher catches associated with lower fishing effort. They recalled childhood experiences where people fished effortlessly in nearshore areas with traditional gear (e.g. spears, traps), and painted vivid pictures of fishermen catching mullet when the fish aggregated in large numbers on Sainte-Marie’s shorelines, before their collapse. As beliefs regarding what healthy fisheries should look like are shaped by fishers’ personal experience, younger fishers could unintentionally accept sparser and potentially less diverse fish populations as a baseline, which might influence support for fishery management [86, 87].

Based on previous research, we expected economically-developed towns to be associated with both conservation-oriented attitudes and willingness to decrease fishing, as fishers living in economically-developed towns might have access to more livelihood options [7, 13, 88]. Instead, sites with evidence of economic development were associated with fishers who continued fishing as a reaction to fishery changes; additionally, there was no correlation between economic development and conservation-oriented attitudes. Fishers’ willingness to exit fisheries was similarly found to be negatively associated with contextual indicators of socioeconomic development in a broad-scale study across five countries in the Western Indian Ocean [89]. Our result extends these findings by suggesting that similar relationships can occur at finer geographic scales. In our study, it is possible that economic development does not improve fishers’ livelihood options; for instance, shops are often family-owned, and the local nightlife may preferentially hire other demographics. Alternatively, access to markets, technology, and in our study the proximity to lagoon (which attract tourists) could also reward fishing in economically developed areas.

Local fishery management

Fishery management restrictions were often suggested to improve fishery health. While perceptions related to potential restrictions were heterogeneous, fishers were collectively more supportive of area-based restrictions than other forms of management. As in many LMMAs in Madagascar, focusing on small strategic areas for management, including restrictions on less selective gear or minimum harvesting sizes, could help protect marine resources and build support for fishery management in the short-term [11, 27, 28]. However, the current condition of the fishery is tightly interlinked with a shift from subsistence to commercial fishery that has altered the scale of the exploitation of marine resources [90]. Consequently, a successful solution will likely need to involve collaborations between the actors contributing to fishery sustainability, including fishers, NGO, and the fishing industry, to develop innovative solutions and achieve shared long-term goals.

Although developing alternative livelihoods could increase fishery sustainability by reducing fishing pressures, our results suggest that it should be approached cautiously in Sainte-Marie. Indeed, we documented persistent fishing behaviours as a response to past changes, as well as a disconnect between fishers’ willingness to decrease fishing and their propensity to suggest management restrictions. Therefore, developing livelihood alternatives would require additional objectives other than increasing fishery sustainability (e.g. risk mitigation, poverty alleviation).

As fishers continue to discuss fishery management rules, good governance will also be key to ensure the success of the initiative. When we conducted this study, fishery management options were often discussed at community meetings. Although these were generally attended by many community members, leadership remained dominated by a select group of powerful individuals that were strongly in support of LMMAs. Acting now to facilitate the effective participation of all fishers, including women, to the elaboration of fishery management plans might be key to ensure the outcomes of fishery management actions [30, 91, 92].

Caveats and future actions

This study presents data collected when project funding had just started and there was no socioeconomic and demographic information for the area. As we did not have any information regarding the number of fishers, sampling efforts were guided by rough estimates, based on consultation with local authorities and communities, of the number of households depending on fisheries in each town. Nonetheless, this study still forms an important assessment of Sainte-Marie’s fisheries as it provides a baseline assessment of fishery practices and perceptions before management restrictions were implemented. The timing of this assessment enables future work to provide insights on how fishery management might influence perceptions, and provides a useful baseline for comparison with other small-scale fisheries in Madagascar and elsewhere.

Our data point to gender differences in fishing practices and perceptions that might have consequences for fishery management; however, our conclusions are hampered by the small number of surveyed women. Indeed, despite the broad definition of fishers and fishing adopted during data collection, traditional authorities and respondents generally directed us to men. Our experience highlights that truly engaging with women requires changing how we collect data [45]. Thus, we encourage improving upon the gender-inclusive approach taken here in future research. In particular, we suggest favouring female-led interviewers as women might be more comfortable talking to the same gender [45]. Using multiple data gathering technique could also be beneficial; for instance, conducting interviews in small, women-only focus groups might allow women to share their perspectives [93, 94]. Finally, alternative to traditional participatory research methods (i.e. interview, focus groups), such as the use of narratives blended with photography, could empower women and encourage them to tell their stories [95].

Although we focused on Sainte-Marie, it is important to gain a better understanding of co-existing fisheries. In particular, fishers highlighted their concern about sea cucumber fishers and shrimp trawlers. A better understanding of the impact of those overlapping fisheries, and how to integrate them into management, is critical because local fishers expect that fishery management will help exclude sea cucumber divers, and to a lesser extend shrimp trawlers, from Sainte-Marie’s waters. Yet, it should be noted that those broader contextual challenges have proven difficult to address with local fishery co-management initiatives [96–98].