Our study is focused on Galicia (NW Spain), where SSF have a key social, historical and economic role in the region (García-Lorenzo et al., 2019). The system is organized into 63 fisher guilds (Confrarías), and is highly diverse in terms of targeted main resources (73 taxa), fishing gears (43 different exploitation techniques and 3,900 vessels), and regulations (García-Lorenzo et al., 2019). Each community in the SSF system is strongly attached to a fisher guild, based on a general top-down governance system. In the fisher guilds, fishers carry out their activity under several management regimes (TURFs, TACs, ITQs and effort-based, gear-based and area-based limitations) (Macho et al., 2013; García-Lorenzo et al., 2019; Pita et al., 2019; Villasante et al., 2019). Tight-knit groups of fishers are created around the SSF activity based on common regulations, fishing gears (métiers; Tzanatos et al., 2006), fishing practices, and the cultural context (Alexander et al., 2018). These groups are associated with the distinctive social identity of fishers and are well known for structuring social interactions (Alexander et al., 2018).

For this work we identified 10 different SSF groups: shellfish gatherers on foot, shellfish gatherers on boat, demersal/specific resources on foot, demersal/specific resources on boat, divers, trap fishers, gillnet fishers, bait fishers, small-scale purse seine fishers and artisanal trawling fishers. Institutional actors, who often play a crucial role in SSF dynamics (Bodin and Crona, 2009), are also integrated in these SSF communities. The number of institutional actors and the roles they take on vary across communities and sometimes overlap within the same person. Here, we identified six main institutional profiles for the Galician SSF communities based on the main functional roles of institutional actors which were corroborated by focus groups (these are: community representative, government representative, secretary, surveillance guard, barefoot advisors, and auction manager) (Macho et al., 2013). The tight cultural attachment of Galician SSF to the marine environment, the complex social context of fisher guilds as the core of SSF communities (García-Lorenzo et al., 2019), and the fact that impacts from climate change are already being recognized (Bode et al., 2008) make the Galician SSF an ideal case study for climate change adaptation research.

We designed a structured survey to be implemented face-to-face where fishers were asked to respond to a set of climate change impact scenarios that affected their income from fishing. Socioeconomic information and indicators of different aspects of adaptive capacity were also gathered, combining Likert scales and multiple-choice questions (Tables S1.2-4). The survey was pre-tested in focus groups in two fishing communities (which were not included in our final sample), with 13 and 12 participants each. The final refined survey was implemented in nine representative fishing communities selected along the Galician coastline (Figure 1). Fishing communities were selected in a stratified manner in order to represent differences in fisher guild size, main target resources and associated fisher groups, geographical location (i.e., covering the south, middle, and northern range of the coastline; and accounting for both the shelter side and open side of the coastline), and presence/lack of protected areas in their seascapes with some restriction for fisheries. A formal agreement with the Galician fisher guild association allowed us to reach the communities and comply with the ethics requirements and guidelines set by H2020 (e.g., the survey protocol included the acceptance of individual fisher guilds to be part of the research project). Five trained interviewers performed the questionnaires face-to-face in the Galician and/or Spanish language during the months of July and August 2018 (in the town, fishing areas, port, guild, and fish market). The average length of the surveys was 45 minutes. All responses were voluntary and anonymous. The final sample of 404 fishers across the 9 communities (Table S1.1) is highly representative of the Galician SSFs system (confidence level higher than 95%), which is roughly composed by 13,500 fishers organized into 63 fisher guilds (Confrarías) (García-Lorenzo et al., 2019).

Following the recommendations resulting from our focus groups, we presented fishers with climate change impacts in the form of scenarios where fishery resources increased or decreased, ultimately affecting the fisher’s income level. Climate impacts on the abundance of primary fishery resources were directly tied to catch levels and associated income from fishing to deter individuals from making predictions about fluctuating prices due to a potential variability in resources resulting from the impacts, which could have biased our results. We presented ten gradual impact scenarios in line with existing research (Daw et al., 2012), with five levels of income increase (15%, 30%, 50%, 70% and 90%) and five levels of income decrease (15%, 30%, 50%, 70% and 90%) associated with climate impacts. Based on information from our focus groups and in line with existing work (Daw et al., 2012; Fedele et al., 2019), we asked fishers to choose between three general adaptation responses for each hypothetical scenario presented in order to classify different types of individual responses to climate change. These were: 1) remain within the SSF as the unique labor activity, 2) diversify their livelihood by engaging in a different activity in addition to fishing, and 3) exit the fishery completely. Respondents were guided to select one single response per climate change scenario (Figure S1.1).

This study is part of a larger project so it has data of a broad range of individual responses but for the purposes of this analysis, which is to uncover the factors that can support the continuation of fishing and fishing culture in the face of climate threats while at the same time alleviating some of the pressure on the fishery theoretically, we focus here on understanding the factors that relate to the livelihood diversification response (#2 described above). In order to identify livelihood diversifiers, we assigned a value of 1 to fishers that, at any given impact level (including increasing or decreasing catch and income), decided to combine their current fishing activity with other external labor outside of the fishery (called ‘diversifiers’ hereafter in the text). Next, a value of 2 was assigned to all fishers who chose to remain in the SSF system and not engage in livelihood diversification outside of the SSF sector across all impact scenarios (referred to as ‘remainers’ hereafter in the text). Finally, those fishers who at any stage chose to exit the SSF completely, and never chose to combine their SSF activities with other external labor activities were coded with a 3 (referred to as ‘exiters’ hereafter in the text). We choose this classification in order to conceptually isolate fishers who chose diversification as a response to climate impacts, as opposed to remaining in the fishery or primarily choosing an exit response. To be clear, while ‘remainers’ never chose to diversify or exit across any of the hypothetical scenarios; ‘diversifiers’ sometimes also chose to exit (depending on the impact scenario they were presented with), and ‘exiters’ never diversify. Importantly however, the vast majority of diversifiers did not choose this response until reaching at least a 70% decline in profits associated with fishing. In contrast, those we classify as ‘exiters’ exhibit demonstrably different exiting behavior, in that most chose to exit in much lower decline scenarios (e.g., 15% decline in profits associated with fishing, Figure S1.2). For further details about these three categories see Figures S1.1-2 and Table S1.2.

Based on recent developments synthesizing an interdisciplinary body of work on adaptive capacity, we developed 22 indicators to capture 7 broad domains of adaptive capacity: (1) assets, (2) flexibility, (3) learning, (4) socio-cognitive constructs, (5) agency, (6) competing concerns and (7) social organization (Table S1.2-4) (Mortreux and Barnett, 2017; Cinner et al., 2018; Cinner and Barnes, 2019).

(1) The assets people have access to or own. People are generally better able to adapt when they can access a diversity of financial, technological and service-related assets (Cinner and Barnes, 2019). Based on existing literature (Cinner et al., 2008; Cinner, 2014; Cinner et al., 2018), we used the number of fishing assets and income as indicators of this domain (Table 1).

(2) The flexibility to make changes. Individuals with more flexibility are better able to adapt to climatic impacts because they are able to access to a broader diversity of potential adaptation options (Cinner et al., 2018). For example, fishing assets (e.g. fishing gear diversity) can enable fishers to adapt to future shocks and adverse trends (Cinner et al., 2018) such as fluctuations of fishing conditions. On the flip side, fishing identity and job attachment have been associated with fishers remaining in fisheries despite deteriorating conditions and diminishing returns (Coulthard and Britton, 2015; Mortreux and Barnett, 2017; Cinner and Barnes, 2019). Age and the dependence on marine resources have also been discussed and identified as important aspects of the diversity adaptive capacity domain in recent work (Barnes et al., 2020; D’agata et al., 2020). In this domain we therefore used fishing identity, job attachment, age, and dependence on marine resources as indicators of flexibility (Table 1). We also accounted for existing livelihood diversity (i.e., the number of complementary income activities aside from fishing) (Cinner and Barnes, 2019; D’agata et al., 2020; Table 1).

(3) Learning to recognize change and assess response strategies. Learning in SESs can help to build awareness of environmental changes and the complex feedbacks and relationships between fishers and marine ecosystems that may underpin, or reinforce them (Cinner and Barnes, 2019). In this sense, SES thinking, and knowledge have been argued to be important determinants driving adaptation (Biggs et al., 2012; Ostrom 2009). Training activities (understood as active engagement) and fishing experience can also be important, as learning is an experiential and experimental process that enables people to frame problems (Shaffril et al., 2016; Shaffril et al., 2019; Cinner and Barnes, 2019). Here, we used SES knowledge, training activities, and fishing experience as indicators of learning (Table 1).

(4) The socio-cognitive constructs that can influence people’s perceptions of risk, as well as the costs and benefits associated with making changes to manage risks (Cinner and Barnes, 2019). In this domain, we measured risk attitude as the degree of subjective risk taken in the SSF labor activity (Mortreux and Barnett, 2017; Cinner and Barnes, 2019; Barnes et al., 2020).

(5) The agency to mobilize adaptive capacities to respond to stressors. This domain refers people`s free choice, empowerment, and self-efficacy in responding to social-environmental changes (Cinner and Barnes, 2019). The household size of fishers is known to affect adaptation strategies (Coulthard and Britton, 2015; Mortreux and Barnett, 2017). For instance, research in farms has associated larger households with the ability to supply surplus labor to non-farm activities as an adaptation practice (Ali and Erenstein, 2017). Also, the presence of children within a household is associated with higher levels of hazard preparedness (Mortreux and Barnett, 2017). In addition, the ability to participate in decision making has been related to agency in previous studies, and affects the capacity to manage prospective situations (Cinner and Barnes, 2019; D’agata et al., 2020). In this respect, we used Household size and Participation in decision making as indicators of agency (Table 1).

(6) Competing concerns capture the effects that different priorities (i.e. the needs to fulfil) might have on adaptive behavior (Mortreux and Barnett, 2017). One way to understand competing concerns is through the lens of gender. Gender can be understood as something people do and, as such, it is a dynamic and contested identity category (Siegelman et al., 2019). In the context of adaptation, existing research suggests that women are less likely to be involved in preparedness for extreme events as they often have multiple responsibilities, and so have less time and energy to commit to household disaster preparation (Mortreux and Barnett, 2017). For this reason and in this context the role of gender is strongly related to the conceptualization of competing concerns, despite its conventional classification as a socio-demographic feature. Unequal power relationships can also contribute to the emergence of competing concerns which influence individual adaptive capacities in the face of changes (Cinner and Barnes, 2019). In this study, we therefore included gender and perceptions of inequality as predictors of competing concerns (Table 1).

(7) The organization of society through social networks of informal and formal relationships. Social organization plays a crucial role in the lives of fishers within our study communities through bonding, bridging, and linking social relationships (Macho et al., 2013; Barnes et al., 2020) that may strongly influence the adaptation process under climate change (Barnes et al., 2017b; Barnes et al., 2020). Bonding ties -ties between people of the same social group- can facilitate cooperation, reciprocity and knowledge co-production that may contribute to adapt SSF while maintaining the dominant structures (Bodin and Crona, 2009; Barnes et al., 2017b; Rockenbauch and Sakdapolrak, 2017). Bridging ties -ties between people of different social groups- are argued to provide knowledge sharing, innovation and access to external resources that may help to initiate transformations (Crona and Bodin, 2006; Bodin and Crona, 2009; Barnes et al., 2017b). And linking ties -ties between people from different levels of an organizational structure- help to leverage ideas beyond the community and promote legitimacy that may facilitate the progress towards transformations (Bodin and Crona, 2009; Barnes et al., 2017b). Here, we address our third research question by following the theoretical foundation laid out in (Barnes et al., 2017b) and testing the presence of bonding, bridging, and linking network structures (Tables 1, A1.4). We do so by distinguishing between (1) communication-based ties, which contribute to information sharing and encourage mutual learning, and thus may facilitate a range of adaptation strategies (Barnes-Mauthe et al., 2014; Gong et al., 2018); and (2) trust-based ties, which may serve to reduce the complexity of a risky situation, and may thus encourage people to adapt in ways that are distinct to communication-based ties (Terpstra, 2011; Table 1).

Further information regarding how each indicator was measured and descriptive statistics of all indicators are provided in Table 1, Appendix S1.1 and Table S1.2-4.

We used a multinomial multilevel logit mixed-effect model (Mbaru and Barnes, 2017) to test the degree to which our livelihood diversification response variable could be explained by our predictors capturing adaptive capacity domains (Cinner et al., 2018; Barnes et al., 2020; D’agata et al., 2020).

Our sample size dropped from 404 to 376 observations due to missing values. The nine communities were included in our model as a random effect (‘Community’, Table 1) to account for potential differences across sites. Results from a variance inflation factor analysis (VIF) demonstrated that there was no collinearity among our indicators (Table S1.5). The model was run with robust variance estimates to account for any potential issues of non-independence related to the manner in which our social and social/social-ecological network structures were conceptualized and calculated (Mbaru and Barnes, 2017; Barnes et al., 2017a). The model was performed in Stata/IC 16.

Participation was voluntary and responses anonymized. Informed consent was obtained from all subjects and/or their legal guardian(s). The experimental protocol was approved by the University of Vigo. Also, a formal agreement with the federal Galician SSF association allow reaching the communities and the acceptance of the fisher guilds to be part of the research project. The whole process has followed the ethical guidelines and requirements set by H2020.

According to our results, we see a pattern where a greater proportion of fishers chose to diversify their livelihoods over other options under extreme negative climate change impact scenarios. This proportion decreases to the 15% negative impact level before levelling out (Figure S1.3).

Under decreasing resource and income scenarios (Figure S1.3), the proportion of diversifiers in our sample increases from 16.83% under the current baseline status to 35.89% under the most extreme resource and income decreasing scenario (accumulative values). Our results are consistent with the work conducted by Galappaththi et al. (2020) where the authors found that livelihood diversification was an important adaptive strategy that indigenous fishing communities used to cope with adverse conditions associated with climate change, such as natural resource scarcity. Coulthard (2008) also noted that as fisheries crises worsen, fishers are increasingly under pressure to seek out new ways of living. Our research contributes to this existing body of work by empirically demonstrating that fishers are likely to choose to diversify their livelihood progressively as scarcity impacts increase, which can help to reduce the risk of livelihood failure (Coulthard, 2009; Ojea et al., 2020).

Our results show that the number of diversifiers drops under positive climate change impact scenarios, as no new fishers decide to diversify (Table S1.6), i.e., the proportion of diversifiers observed for our baseline scenario (16.83%) remains constant across all positive impact scenarios (Figure S1.3). Thus, under hypothetical scenarios of increasing biological resources, we found that fishers neither change their livelihoods nor increase their livelihood diversification strategy. Existing research regarding the role of livelihood diversification under scenarios where resources or economic profits increase is largely anecdotical, though this finding does contradict an existing study in which an increase in forest land cover at Nqabara on South Africa’s Wild Coast led people to shift from large cultivated fields, which had become overgrown, to smaller gardens (Fabricius et al., 2007). In other words, changes in livelihoods were positively associated with an increase in natural resource abundance (Fabricius et al., 2007). Our results thus suggest that when faced with climate impacts, the transformative behavior of small-scale fishers may differ from that of actors in other natural resource-dependent systems; yet additional research is needed in order to confirm this hypothesis. Other possible drivers of the adaptive behavior we studied here include environmental and contextual conditions, which may also lead to different adaptation strategies in different places and sectors (e.g., fishers and farmers).

Overall, 146 (36.14%) fishers stated they would choose to diversify their livelihood by finding work outside the SSF at least once under the set of 10 climate impact scenarios, while 258 (63.86%) fishers never choose this strategy (112 remainers and 146 exiters) (Table S1.2). The somewhat low proportion of livelihood diversification responses overall for the Galician case study may be explained by several potential factors. First, fishers may not feel that they have suitable livelihood alternatives, as existing research suggests that fishers experiencing impacts on livelihoods will feel motivated to diversify and even exit the fishery when suitable alternative incomes are offered (Muallil et al., 2011). Second, fishing is often considered more than just an occupation, and fishers have strong attachments to their work (Bavinck et al., 2012). Indeed, research conducted across several countries revealed that fishers enjoy their occupation and are not often interested in alternative employment (even if catches and associated incomes fall), which tends to be contrary to the assumptions underpinning many fisheries policies (Monnereau and Pollnac, 2012). Although this second explanation could be plausible, the variables we use to measure job attachment and fisher identity (job attachment and fisher identity) do not show any level of significance in our results.

Results of the multinomial multilevel logit mixed-effect models for diversifiers are depicted in Figure 2. Generally, our findings show that flexibility enables livelihood diversification, however different aspects of learning, competing concerns and social organization play both enabling and constraining roles on this transformative process (Figure 2). Overall, compared to those who remain in the fishery without engaging in any form of livelihood diversification (remainers), those who choose to transform their livelihoods by diversifying are less likely to perceive inequality in the system, have greater livelihood diversity to begin with, depend on a greater diversity of marine species, have less bridging trust, less bonding communication and more bridging communication. Comparing diversifiers to those who choose to exit the fishery (exiters), diversifiers have greater (existing) livelihood diversity, SES knowledge, less fishing experience and lower levels of inequality perception in the system. In this comparison, fishers categorized as woman and those with trust-based linking ties were also more likely to diversify livelihoods.

In the following sections, we interpret these results in light of the research questions, looking first at the effect of adaptive capacity indicators on diversification, and second on the role of specific social-ecological network structures comprised mainly within the social organization domain.

Critically, we find that having higher levels of understanding of the SSF as a SES (SES Knowledge) is significantly, positively (p-value = 0.021) related to decisions to diversify livelihoods rather than exit the fishery under climate impact scenarios. This finding is in line with theoretical arguments describing SES thinking and knowledge as critical elements of resilience (Biggs et al., 2012; Ostrom 2009). Cinner and Barnes (2019) argue that learning reflects people’s capacity to generate, absorb, and process new information about climate change and adaptation options; and ways to live with, and manage, uncertainty. Learning may contribute to building robust decision-making capacity under adverse conditions, allowing people to engage in complementary livelihoods to reduce risk if necessary. Despite this rich literature describing the key role of learning and SES thinking to navigate uncertainties (Biggs et al., 2012), our results are the novel in empirically demonstrating the enabling role of fisher’s SES knowledge in livelihood diversification under climate impact scenarios.

Our results also show that Fishing experience contributes negatively (p-value = 0.002) to diversification decisions when compared to those who primarily chose to exit (Figure 2). In other words, the more time a fisher participates in a SSF, the less likely they are to diversify as opposed to exit. This result is contrary to what we might expect since the existing literature suggests that fishing experience may be related to a strong occupational attachment, which we might expect to lead fishers to choose to diversify rather than simply exiting fisheries under climate impact scenarios (Shaffril et al., 2016; Shaffril et al., 2019). A simple explanation might be that more experienced fishers (who are often older) prefer to leave SSFs rather than to diversify livelihoods under impact scenarios because they may be expecting to retire. In fact, age and fishing experience (tested for multicollinearity which showed no issues between these two variables) were positively correlated in our case (R= 0.62, p < 0.000). Although age was not significant in our model, the age of exiters is significantly higher than the age of diversifiers (Table S1.8). Another potential explanation might be that the more experienced small-scale fishers are, the more knowledge they may have regarding climate-related consequences on SSFs (Mortreux and Barnett, 2017), which may cause them to see livelihood diversification as an unsuitable way to fulfill their needs and sustain their life into the future. This rationale would be consistent with Beaudreau et al. (2019), who found that contrary to portfolio theory, specializing (rather than generalizing, which would result from diversifying) was associated with stronger benefits for fishers (Beaudreau et al., 2019).

Flexibility also played an important role in livelihood diversification decisions (p-value = 0.000 compared to remainers and p-value = 0.001 compared to exiters), consistent with arguments that individuals with more flexibility may be better positioned to adapt to climatic impacts (Cinner et al., 2018). First, we found that those who are already diversifying livelihoods (Livelihood diversity variable) will be more likely to further diversify livelihoods under climate change scenarios rather than remain in the fishery or exit. Fishers that have already diversified their livelihoods may be more familiar with the diversification process and can apply those specific skills under future uncertainties. While this sounds intuitive, it actually stands in contrast to the results from previous studies in SSFs that found that having more occupations (i.e., greater livelihood diversity) increased the likelihood of exiting the fishery (Cinner, 2014). Empirical examples testing the role of current livelihood diversification on fisher’s adaptation decisions under uncertainty are scarce; thus, more research may be needed to understand the factors that influence fisher’s decisions to diversify as opposed to exit under different levels of impact or risk. We also found that fishing a more diverse portfolio of fishery resources (Dependence on main marine resources) was positively related to livelihood diversification compared to those whose mainly chose to remain in the fishery. We describe this result in further detail in section The role of network structures on livelihood diversification.

We find competing concerns as the third adaptive capacity domain that has a strong relationship with diversification. Firstly, compared to both remainers and exiters, Inequality is negatively related to livelihood diversification decisions (p-value of 0.037 and 0.025, respectively; Figure 2). This means that the more the fishers feel they are in an unequal situation within the SSF community, the less likely they are to diversify. This novel result is consistent with Maslow’s hierarchy of needs theory which has been adapted to fisheries realms (Cinner and Pollnac, 2004). In Maslow’s hierarchy, Inequality may reflect unfulfilled social needs such as social belonging and acceptance. According to the theory, these unfulfilled needs prevent people from focusing on reaching higher levels of the pyramid, such as esteem, cognitive and self-actualization (Cinner and Pollnac, 2004). Esteem, cognitive and self-actualization are strongly related to agency and other domains of the adaptive capacity framework which may be crucial for livelihood diversification (Cinner and Barnes, 2019). In the same line, the psychology literature describes that people have a limited capacity to worry, such that increases in worry about one issue in life will lead to a decrease in worry about other issues (Mortreux and Barnett, 2017). In this aspect, our finding is consistent with the literature that suggests that the presence of competing concerns significantly impedes adaptation, in this case through livelihood diversification, even when adaptive capacity is high (Mortreux and Barnett, 2017). Indeed, this perceived inequality could be an indicator of the existence of social-ecological traps (Cinner, 2011), where for instance, oppressing powers create a situation where marginalized groups and individuals remain locked in maladaptation, further eroding their wellbeing (Blythe et al., 2018). To solve these conflicts, deliberative processes are argued to be key (Ostrom 2009). Overall, our results suggest that equity considerations may be critical to facilitating climate adaptation and transformation in SSFs.

Secondly, we find that Gender is strongly related to livelihood diversification decisions under climate impact scenarios; i.e., compared to those who largely chose exit responses, women are more likely to choose livelihood diversification responses (p-value = 0.007) (Figure 2). These results are consistent with the literature that describes gender as an important predictor of attitudes towards natural disasters (Ho et al., 2008; Haynes et al., 2010; Eriksen et al., 2010). Within fisheries, women have been shown to play a key role in both “endogenous strategies” (such as maintaining nets or housekeeping) and “exogenous strategies” (such as working outside of fisheries to receive a supplement income) to cover the needs of the household and contribute to buffering impacts (Coulthard and Britton, 2015). This role may contribute to vulnerabilities when faced with risks (Coulthard, 2012; Coulthard and Britton, 2015; Kawarazuka et al., 2017), e.g., women may be less able to migrate out of fisheries completely, even under extreme climate impacts (Kawarazuka et al., 2017) (Cinner, 2011). However, SSF are often a way of life; thus, fishers that are able to find ways to remain in the fishery while still adapting or transforming in the face of impacts may be better off in terms of their wellbeing (Shaffril et al., 2016; Shaffril et al., 2019). Our findings demonstrate the gendered nature of climate change adaptation in SSF. With the aim to improve this research line, we encourage further research to go beyond gender and integrate the LGTBIQA+ research in order to better understand the social diversity of SSF (Dominey-howes et al., 2014; Resurrección et al., 2019). This will help us to fully disentangle the outcomes of diversification decisions and better identify marginalized groups and inequities, as well as good practices to combat them, when considering climate change adaptation and transformation strategies (Dominey-howes et al., 2014; Resurrección et al., 2019).

Though existing research has shown that assets, agency and socio-cognitive constructs can be important for adaptation and transformation (Barnes et al., 2020; D’agata et al., 2020; Mortreux and Barnett, 2017); we did not find evidence that they were related to livelihood diversification in our analysis of Galician SSFs under climate impact scenarios. Assets in particular are commonly considered to be one of the most important factors in building adaptive capacity (Yohe and Tol, 2002; Smit and Wandel, 2006), and as such are the focus of many adaptation programs (Lemos, 2007; Cinner et al., 2018). Here, we only included two predictors of assets: fishing assets and income. One possible reason that might explain why income and fishing assets was not significant in our models for predicting livelihood diversification responses is the coexisting conditions of a “growth-based western economy country” context, whereby every fisher is maintained above a minimum level of income and fishing assets, which may dampen the effect of income level. It is possible that other types of assets that we did not measure may have also been important, such as additional material possessions, savings, or access to credit. Similarly, we were only able to capture a limited number of predictors of agency and socio-cognitive constructs. Future research should expand on our efforts to develop a better understanding of how these domains may shape adaptive decisions.

We find five network structures related to livelihood diversification. We explain these findings in five main key points. First, diversifiers show a significantly lower level of Bonding communication (p-value = 0.007) than remainers (Figure 2). In other words, fishers who communicate within their fishing groups are more likely to remain in the fishery than to transform by diversifying their livelihood. This is consistent with the literature that explains that bonding ties promote knowledge co-production, trust, reciprocity, and cooperation, all of which can maintain the dominant structures and function of a SES (Bodin and Crona, 2009; Berardo, 2014; Rockenbauch and Sakdapolrak, 2017). Specifically, bonding ties can provide a sense of identity and belonging, social support (in time of crisis), shared vision and mutual learning which may positively influence fishers’ perceptions of remaining in fisheries through common agreements and governance practices that facilitate the adoption of technologies and adaptation strategies (Barnes-Mauthe et al., 2014; Barnes Truelove et al., 2015; Gong et al., 2018; Barnes et al., 2019; Wilson et al., 2020).

In contrast to our results on bonding communication, Bridging communication has a significantly positive effect (p-value = 0.033) on livelihood diversification choices (Figure 2). In other words, fishers who communicate with different people are more likely to diversify their livelihood than to remain in SSF as their only labor activity. This is consistent with the literature that explains that bridging ties between people of different groups can provide access to a diversity of external resources, enhance innovation, promote creativity, and provide new perspectives that can be crucial in times of crisis (Newman and Dale, 2005; Crona and Bodin, 2006; Bodin and Crona, 2009; Rockenbauch and Sakdapolrak, 2017). These features can enable the development of new perspectives that initiate livelihood transformations (Crona and Bodin, 2006; Bodin et al., 2006; Bodin and Crona, 2009; Rockenbauch and Sakdapolrak, 2017; Barnes et al., 2017b). Aligned with our results, Zhang et al., 2020 find in their research on the agriculture sector that people who maintain communicative bridging ties also display more progressive practices to help manage uncertainty and change.

Third, the majority of research on social networks and adaptation/transformation has focused on communication ties. Critically, in our model Bridging trust has a significant (p-value = 0.088) negative effect (Figure 2). In other words, those with high levels of Bridging trust ties are more likely to remain in the SSF rather than transform by diversifying their livelihood. Our results are consistent with literature that argues that a high level of trustful relationships may pose a barrier to transformation due to constraining social norms of mental lock-ins (Barnes et al., 2017b). In this line, this network structure combines the effect of trust (mental lock-ins) and the effect of bridging ties (the connection among different groups within SSFs), which may lead to a “we” (small-scale fishers) versus “them” (non-small-scale fishers) mentality (Dressel et al., 2020). Also, in order to reduce the complexity of the situation, trust-based ties may influence fisher’s risk judgments (Terpstra, 2011; Zhang et al., 2020). Specifically, when trust is high, individual risk perception may be low, potentially causing people to perceive livelihood diversification as unnecessary (Terpstra, 2011; Wilson et al., 2020; Bodin et al., 2020). To date, little research has focused on the different types of bridging relationships and transformation in SES. Our findings demonstrate that this distinction is important, since communication-based bridging ties are positively related to livelihood diversification, whereas trust-based bridging ties are negatively related to livelihood diversification.

Fourth, Linking trust has a significant (p-value = 0.092) positive effect on livelihood diversification behavior compared to exiting behavior (Figure 2). In other words, fishers who trust institutional actors of the SSF community are more likely to diversify livelihoods in times of climate-induced impacts, rather than exit SSF. This is probably because (1) linking ties provide access to resources and knowledge beyond the levels of the community and can promote legitimacy (Newman and Dale, 2005; Bodin and Crona, 2009; Barnes et al., 2017b), (2) trust-based linking ties may indicate trust in leaders to manage the risks of remaining in the fishery (Terpstra, 2011), and (3) fishers may feel empowered because linking ties often enable fishers to have some agency over fishery decisions that affect them (Marín et al., 2012). Our results are consistent with previous research conducted on a Chilean coastal benthic resources co-management system which showed that the investment of small-scale fisher organizations in linking social relationships was associated with more diversified livelihoods (Marín et al., 2012).

Finally, Dependence on main marine resources has a significant, (p-value = 0.064) and positive effect on livelihood diversification decisions (Figure 2) compared to remaining in the SSF. In other words, fishers who fish a greater diversity of marine species are more likely to transform by diversifying their livelihood than to remain in SSF as the only labor activity when faced with climate impacts. This is consistent with the literature that explains that social-ecological ties represent the environmental knowledge and capacity acquired from the connection with different target biological marine species that may enable fishers to detect environmental changes and anticipate future change scenarios (Barnes et al., 2017b; Barnes et al., 2020). As Barnes et al., (2017b) mention, resourceful actors linked to multiple resources -targeting different main marine species- can facilitate experimentation while reducing risk, which is critical in the initial phase of transformation. Our findings demonstrating a significant positive relationship between social-ecological networks and individual livelihood diversification strategies under climate-driven impacts adds novel insight to recent advances in the SSF adaptation literature (e.g. Barnes et al., 2020).

The role of timing in climatic crises and their influence in fishers’ decision making is an important point to consider. Although we did not explicitly study the time dimension, the extent of climate impacts on income could represent the extremity or acuteness of an unfolding climate crisis, with lower income impacts potentially representing a crisis unfolding more slowly (precautionary steps and proactive behaviour) vs. large income declines being connected to extreme events (reactive behavior). Further research is needed to explore the role of the timing and worsening conditions on adaptation, which may play an important role in how people respond.