Climate change, biological invasions, and the shifting distribution of Mediterranean fishes: A large‐scale survey based on local ecological knowledge

Climate change and biological invasions are rapidly reshuffling species distribution, restructuring the biological communities of many ecosystems worldwide. Tracking these transformations in the marine environment is crucial, but our understanding of climate change effects and invasive species dynamics is often hampered by the practical challenge of surveying large geographical areas. Here, we focus on the Mediterranean Sea, a hot spot for climate change and biological invasions to investigate recent spatiotemporal changes in fish abundances and distribution. To this end, we accessed the local ecological knowledge (LEK) of small‐scale and recreational fishers, reconstructing the dynamics of fish perceived as “new” or increasing in different fishing areas. Over 500 fishers across 95 locations and nine different countries were interviewed, and semiquantitative information on yearly changes in species abundance was collected. Overall, 75 species were mentioned by the respondents, mostly warm‐adapted species of both native and exotic origin. Respondents belonging to the same biogeographic sectors described coherent spatial and temporal patterns, and gradients along latitudinal and longitudinal axes were revealed. This information provides a more complete understanding of the shifting distribution of Mediterranean fishes and it also demonstrates that adequately structured LEK methodology might be applied successfully beyond the local scale, across national borders and jurisdictions. Acknowledging this potential through macroregional coordination could pave the way for future large‐scale aggregations of individual observations, increasing our potential for integrated monitoring and conservation planning at the regional or even global level. This might help local communities to better understand, manage, and adapt to the ongoing biotic transformations driven by climate change and biological invaders.

dance was collected. Overall, 75 species were mentioned by the respondents, mostly warm-adapted species of both native and exotic origin. Respondents belonging to the same biogeographic sectors described coherent spatial and temporal patterns, and gradients along latitudinal and longitudinal axes were revealed. This information provides a more complete understanding of the shifting distribution of Mediterranean fishes and it also demonstrates that adequately structured LEK methodology might be applied successfully beyond the local scale, across national borders and jurisdictions. Acknowledging this potential through macroregional coordination could pave the way for future large-scale aggregations of individual observations, increasing our potential for integrated monitoring and conservation planning at the regional or even global level. This might help local communities to better understand, manage, and

| INTRODUC TI ON
The redistribution of Earth's species is among the most evident consequences of global warming (Parmesan & Yohe, 2003;Poloczanska et al., 2016) and a critical aspect for the health of both natural ecosystems and human populations worldwide . These changes are usually greater for marine environments, because of their high environmental connectivity (Burrows et al., 2011) and because of the pivotal role of water temperatures, which strongly influence growth, survival, and reproduction in marine animals (Crozier & Hutchings, 2014;Reusch, 2014). In fact, even apparently modest changes in water temperature might trigger a rapid cascade of multiple pressures over marine organisms. Some species, unable to cope with these environmental alterations, or benefit from them, may change their abundances accordingly. However, mobile marine organisms also have another option: they can move to new areas where they were formerly absent (Cheung et al., 2009;Fogarty, Burrows, Pecl, Robinson, & Poloczanska, 2017). These two dynamics are not mutually exclusive, as they can be considered as different behavioral and demographic responses that might coexist in the same species or population.
Despite the magnitude of these changes and their relevance for conservation and adaptation policy (Givan, Parravicini, Kulbicki, & Belmaker, 2017;Marras et al., 2015), observational studies are often fragmented in space (Elmendorf et al., 2015) and methodologically heterogeneous (Coll et al., 2010). This also applies to the northward expansions of warmwater species, a phenomenon that has been mostly described in the northwestern sectors of the Mediterranean basin, probably due to the uneven distribution of research efforts (Boero et al., 2008;Lejeusne, Chevaldonné, Pergent-Martini, Boudouresque, & Pérez, 2010;Marbà, Jordà, Agustí, Girard, & Duarte, 2015;Sabatés, Martín, & Raya, 2012). This fragmentation, together with the lack of coherent depictions of change, hampers the availability of reliable information to stakeholders and decision-makers (Grafton, 2010;Pauly & Zeller, 2016). Indeed, in light of profound impacts that have already affected both people and the ecosystems they depend on, many national and transnational authorities and agencies are engaged in efforts to build adaptive capacity, seeking reliable information to enable people to anticipate and appropriately respond to the ongoing change (Coulthard, 2012). This explains the growing need of integrated monitoring and assessment systems to capture the ongoing transformations of marine ecosystems (including the effects of a changing climate) and to bring them into the policy agendas (Creighton, Hobday, Lockwood, & Pecl, 2016). Certainly, our observational potential grew steadily during the last few years and increasing efforts are devoted to conceive global observation systems for up-to-date information on the state of biodiversity and the threats it faces (Tittensor et al., 2014). To achieve this, the use of standardized and cost-effective procedures is needed to underpin a large-scale observation strategy that can accommodate countries across a range of baseline knowledge levels and capabilities (Bélisle, Asselin, LeBlanc, & Gauthier, 2018;Latombe et al., 2017). These are key principles for collecting and integrating information from stakeholders across national boundaries. In this, fishers are a particularly interesting group of stakeholders, as they spend a considerable proportion of their lives in close contact with the marine environment and they become familiar with local species. Therefore, their personal experience gained through individuals' observations over their lifetimes can provide precious complementary information about marine communities and be used to set effective monitoring practices. Yet, accessing this expert knowledge (hereafter referred as local ecological knowledge or LEK) is offering new opportunities to Mediterranean research Azzurro, Moschella, & Maynou, 2011;Bastari, Beccacece, Ferretti, Micheli, & Cerrano, 2017;Coll et al., 2014;Damalas et al., 2015;Mavruk, Saygu, Bengil, Alan, & Azzurro, 2018), providing new opportunities to overcome practical and budgetary constraint, especially in poorly studied areas.
Here, we accessed the knowledge of Mediterranean fishers, to reconstruct changes in fish distribution and abundance, altogether with their related spatial and temporal dynamics. We did so by: 1. Compiling a dataset of species that were perceived as increasing or new by respondents (hereafter referred to as increasing species); 2. Using this multivariate information to explore the structure of perceived change across different subsectors of the Mediterranean

Sea;
3. Testing for the effect of spatial gradients on the overall number of increasing species; 4. Exploring the spatiotemporal evolution of increasing species.

| Fishers' interviews
Drawing on the methodology conceived within a pilot experience (Azzurro et al., 2011) and according to the procedure described by Garrabou, Bensoussan, and Azzurro (2018), we used a semistructured questionnaire to reconstruct changes in distribution and abundance of Mediterranean fishes.
Knowledgeable small-scale fishers with more than 10 years of experience were identified and selected by each local research team and individual face-to-face interviews were realized according to a standard protocol. Respondents were asked to mention the species that increased in abundance or were perceived as "new" (i.e., never observed before) in their fishing areas. For each of these species, qualitative ranking of historical abundances was expressed along a yearly timeline and according to six categories This large spatial coverage was made possible through a collective and coordinated effort based on the engagement of an international team of researchers well connected with local fishery communities.
The methodological transfer to the participating researchers was supported, from 2012 to 2016, by five training sessions carried out in Tunisia, Montenegro, Albania, Croatia, and Italy. Training included both theoretical lessons and joint field surveys made in collaboration with local fishers. Attendants were guided in performing standardized interviews and advised on how to reduce potential biases, such as the ones related to taxonomical identification and "memory recall" bias (Coughlin, 1990). Interviews were realized between 2009 and 2016 by local researchers in local languages (Albanian, Arabic, Croatian, Greek, Italian, Montenegrin, and Turkish). The LEK protocol is currently applied in other Mediterranean countries, such as Libya, Spain, and France and adopted by five Mediterranean marine protected areas generating new data, which were not included in the present study.

| Sample characteristics
A total of 513 Mediterranean fishers with more than 10 years of experience were selected and successfully interviewed. Their age ranged from 28 to 87 years (mean ± SD: 48 ± 11). Their cumulative working experience accounted for a total of 15,030 years of observations at sea. Overall, 59% of respondents were represented by professional fishers and 38% by recreational ones. Gillnets were the most common used gear among professionals (48%), followed by longlines (26%), traps (9%), purse (8%), and other gears (9%).

| Statistical approach
Based on available literature (Azzurro, 2008;Golani et al., 2018)  Autosimilarity curves are adopted in community ecology to see if sample size is suitable to detect all the species within a community (Schneck & Melo, 2010). A curve is calculated by iteratively computing average resemblance values between randomly selected samples from a dataset. When resemblance attains an asymptote, sample size is deemed to represent a whole community. In this research, we regarded interviews as ecological samples. Therefore, autosimilarity curves told us whether our sampling in the various areas of the Mediterranean captured fisher's consensus about increasing species. We fit separate curves for NIS, NES, and OIS.
Finally, to see the extent to which changes in fish communities were reflected in fisher's knowledge, (d) we modeled the effect of latitude and longitude over the total number of increasing species and over the number of increasing NES, NIS, and OIS, through generalized additive modeling (Guisan, Edwards, & Hastie, 2002;Hastie & Tibshirani, 1990;Wood, 2017a;Wood, Pya, & Säfken, 2016). To account for heterogeneity in sampling effort, we used the total number of interviews collected at each location as an offset. We chose a spline-based penalized likelihood estimator, with a fixed number of knots (k = 6), that was deemed large enough to avoid overfitting and Wald chi-square statistics was adopted to test for the significance of smooth terms (Wood, 2013).
Spatiotemporal changes in fish abundances were analyzed through breakpoint analyses of the historical time series of perceived abundances of the two most frequently cited NES and NIS species. We determined the year at which each species-specific time series indicated a significant change in the perceived abundance (breakpoint) by using a binary segmentation method assuming a Poisson distribution of the data (Killick & Eckley, 2014). To quantify the intensity of this break, we also determined its jump, defined as the difference between the perceived abundance before and after the breakpoint. Since the breakpoint analysis was not sensitive in detecting the exact year of arrival of the "new" species, we also extracted from each species-specific time series the year of perceived arrival, which corresponded to the year at which the perceived abundance changed from 0 (absence) to any of the other scores (i.e., 1-5). Then, we explored the effect of latitude and longitude over the year of break, the jump, and the year of arrival, through another set of GAM with a gaussian distribution of the error. We implemented six models for each species using latitude and longitude as smoothing terms for the three variables (year of break, jump, and year of arrival). In all cases, the total number of interviews collected at each latitude and longitude was used as offset to account for different sampling efforts. Then, we used spline-based penalized likelihood estimators and a number of fixed knots (n = 7) and F statistics was used to assess the significance of smooth terms (Wood, 2013).

| RE SULTS
Mediterranean fishers, with their varying cultural and political settings, were proved a fertile ground where to explore LEK on changes in fish diversity and abundance. In most of the cases, respondents were interested about the research questions, glad to share information with the researchers, and generally pleased to be regarded as experts. What most participants pointed out in their narratives was the rapid and dramatic ecological change, and the reconstruc- This kind of ecological memories are often linked to an emotional dimension (in this case a new species never seen before) and can be associated with personal histories (in this case the birth of the second child). This helped tracing back-specific temporal trends through the diagramming approach, transforming fisher's narratives in temporal series.

| Species perceived as increasing in abundance or new in respondent's fishing areas
Overall, 423 fishers (82%) told us that at least one species increased in abundance or appeared as new in their fishing area, for a total of 886 observations across 75 taxa. These included a number of 13 NIS (21% of citations), 20 NES (64% of citations), and other 42 OIS (15% of citations). A complete list of species is available in Figure S1.

The invasive Lagocephalus sceleratus and Fistularia commersonii
were the most cited NIS (31% and 34% of total observations, respectively, see Figure S1), while Pomatomus saltatrix and Sphyraena viridensis were the most cited NES (30% and 15% of total observations, respectively, see Figure S1). Finally, Sparus aurata, Synodus saurus, and Thunnus thynnus were the most cited OIS (16%, 10%, and 9% of total observations, respectively, see Figure S1). Some of the autosimilarity curves, based on the Bray-Curtis similarity index, reached an asymptote (Figure 2a), indicating that respondents strongly agreed on the increase of a specific group of species. This was observed for NES in all the subsectors of the Mediterranean but the Levantine, and for OIS, like Sparus aurata, in the Tyrrhenian and the Adriatic Sea (see Table S1). Respondents belonging to the same geographical subsectors generally provided coherent information about NIS, NES, and OIS, when interviews were collected from the same geographical sector (e.g., the Tyrrhenian Sea). On the contrary, significant differences can be highlighted for the group NIS, when distant areas are compared (e.g., Tyrrhenian vs. Levantine Sea; Figure 2b). Similarity percentages expressed through the Bray-Curtis index (Table S1) showed the species which explained the most observed similarity between responses. For example, respondents from the Adriatic and Tyrrhenian areas provided similar depictions of change, because they agreed over the increase of P. saltatrix as the most important species accounting for the observed intragroup similarity (Table S1). Intragroup similarity, in other subsectors like the Tyrrhenian, the North Aegean, or the Strait of Sicily, was explained by a wider group of species (Table S1). A complete table of the var-ious NIS, NES, and OIS cited as increasing in the various subsectors is available in Table S2.

| Spatial gradients in the overall number of increasing species
Latitude and longitude explained 33.5% of the deviance in the total number of species mentioned by the respondents (R 2 = 0.54; UMBRE = 0.267; see also Table S3). The number of cited NIS showed a significant and linear decrease along a northward gradient, with higher number of NIS at lower latitudes (Figure 3). On the contrary, no effect of longitude was highlighted (p > 0.05).
Concerning OIS, these species did not show any clear, nor significant (p > 0.05), latitudinal pattern. On the contrary, their number significantly decreased from lower to higher longitudes (p < 0.001).
Finally, the number of NES increased between 33 and 40 degrees of latitude, and remained stable at higher latitudes ( Figure 3, Table S3).
A significant (p < 0.001) smooth effect of longitude with constant values up to 23 degrees, followed by a steep drop was also observed ( Figure 3, Table S3).

| D ISCUSS I ON
In this research, we used for the first time LEK to reconstruct distributional changes in species across an entire geographical region, the Mediterranean Sea. Our approach responds to the idea of collecting a minimum set of essential variables, which can be used to ensure effective collaboration among countries and tangible information on a  (Marbà et al., 2015). Northward spreads were extremely obvious for species such as the bluefish, P. saltatrix, which was reported to positively respond to seawater warming in both the northwestern Mediterranean (Sabatés et al., 2012) and in the Atlantic Ocean (Callihan, Takata, Woodland, & Secor, 2008). Similar to the bluefish, other native and exotic warm-adapted species might have taken the advantage of changing environmental conditions (Lasram & Mouillot, 2009)  Lessepsians are typically very common in the eastern Mediterranean sectors but may be rare or even absent in other geographical sectors, such as the eastern Adriatic, the North Aegean, and the most of the Northwestern Mediterranean Sea (Golani et al., 2018). Here, GAM highlighted a latitudinal and a longitudinal effect over the number of reported NIS, and the change in the NIS pool across longitude reflects the geographical structure of the Lessepsian bioinvasion, whose importance progressively increases when we move to the east and to the south of the basin (Golani et al., 2018). Fisher's observations also illustrated that the Lessepsian phenomenon, once confined to the eastern sectors of the Mediterranean, has rapidly progressed to the west, vanishing the boundaries of the so-called "Lessepsian Province" (Por, 1990). This was particularly clear for the Sicily strait, which was historically considered as an insurmountable barrier to the dispersion of Red Sea fishes and it is now colonized by species such as rabbitfishes (Siganus luridus) and cornetfish (F. commersonii).
While the picture provided by NES and NIS shows coherent responses over entire geographical subsectors, confirming the influence of large-scale drivers, the increase of the remaining species (OIS) can be mostly attributed to local causes, or to the finding of rare/uncommon species perceived as "new" by the respondents. and widespread mariculture and associated unintentional escapees (Dempster et al., 2018), which might act as inadvertent but continuous restocking of this species over large areas of the basin.
Spatial patterns are well illustrated by the nMDS (Figure 3) and the plotted distances of reported observations show that respondents from different subsectors of the Mediterranean might hold different experiences. For example, Levantine and Adriatic fishers did not overlap in terms of cited species, and this is primarily explained by the great differences held by these sectors in terms of community composition.

| Temporal dynamics and their spatial variation
The breakpoint analysis identified critical changes in both spatial and temporal dynamics of cited species. For example, the arrival of F. commersonii was extremely sudden at lower latitudes around the year 2000 and then positively increased toward 2010 with lower strength, matching the strength and rates of its invasion history, as reconstructed through published observations (Azzurro, Soto, Garofalo, & Maynou, 2013). The expansion of P. saltatrix was mostly reported from the northwest of the Mediterranean Sea, while any significant breaks and/or arrivals were recorded in the southeast sectors of the Mediterranean, where the species historically occurs (Sabatés et al., 2012).
Overall, the first evidence on the northward expansion of warmwater species was provided in the 1990s (e.g., Bianchi, 2007;Bianchi et al., 2012;Francour, Boudouresque, Harmelin, Harmelin-Vivien, & Quignard, 1994), while a clear increase in sea temperature and important changes in the water circulation of the Mediterranean Sea are visible since the 1980s (Boero et al., 2008). The critical changes illustrated by our temporal reconstructions and breakpoints confirm and describe the increase of warmwater species at higher latitudes. For example, the dynamics of the bluespotted cornetfish F. commersonii agrees with the onset of its Mediterranean invasion (in 2000), and most interestingly, the strength of the breaks (jumps) was particularly great at higher latitudes, mirroring the rapid demographical explosion of this species in the easternmost sectors of the Mediterranean (Golani et al., 2018). A similar pattern of rapid population explosions was reconstructed for the silver-cheeked toadfish L. sceleratus, which showed very strong breaks in the easternmost sectors of the Mediterranean, since 2003, hence, immediately after its detection.

| Strengths and weaknesses of a large-scale LEK survey
The non-structured approach of our interviews allowed each respondent to spontaneously mention "new" or increasing species in each fishing area, so each interview may be considered as an independent replicate in our design. The high degree of coherence among respondents from the same geographical subsector improved the confidence in the fact that trends reflect real patterns in the environment, with promising outcomes for largescale investigations. Indeed, the logic of focusing on a regional change is analogous to that for global or climate changes itself. As highlighted by (Parmesan & Yohe, 2003), surveying for large-scale fingerprints does not require that any single species is driven by a large-scale determinant with 100% certitude. Rather, it seeks some defined level of confidence in the whole signal. Also, the extent of our geographical scale makes our findings relatively robust against cognitive biases, framing effects, and memory recall issues, that are likely to affect detailed and punctual records in space and time, rather than overall, coarse, estimates (Vaske, 2008). This also apply to other potential sources of variability, such as the attitude of respondents, their different fishing gears (Azzurro et al., 2011), and the limited access to particular depths or areas (e.g., Beaudreau & Levin, 2014). Certainly, the influence of factors such as climate change and fisheries on the observed dynamics was not specifically tested in this study. In this regard, we might note that only a restricted subset of Mediterranean NIS was mentioned, representing only the most recent invasions.
Other invaders were not cited by the respondents, because they were not perceived as new or increasing in their fishing areas.

| Concluding remarks
The whole Mediterranean is rapidly changing its biotic identity and accessing the knowledge of local fishers provided us with an improved understanding of the recent spatiotemporal dynamics of species "on the move," mainly represented here by warm-adapted fishes expanding across the basin. The resulting picture helps to fully appreciate the regional dimension of species redistributions, which will leave "winners" and "losers" in their wake . Merging local efforts together, we build practical bridges to deal with these complex and large-scale transformations. To this regard, we must consider the inherent value of LEK on its own, and not only as sole ecological information. Indeed, individuals who inhabit sites of change might possess deeper readings of place and undisclosed capacities for interpretation (Mustonen, 2014) and adaptation (Berkes, Colding, & Folke, 2000). Therefore, the valorization of their knowledge is expected to reinforce our potential for adaptive ecosystem-based management, improving the operability of future actions in the real world (CIESM, 2018). The importance of this interplay, between ecological and social aspects, has been largely recognized by the scientific community (Allen, Fontaine, Pope, & Garmestani, 2011;Bennett et al., 2017;Berkes, 2004; and it is considered as a key ingredient to support robust and effective conservation policies in the Mediterranean region (Katsanevakis et al., 2017). In conclusion, advancing the use of LEK across large geographical scales allows bringing together the voices of people from different countries, ultimately preparing for a world of global ecological change. We believe that this beneficial partnership, which was here demonstrated to provide tangible results at the regional scale, could be extended to assessments at the global scale, if properly designed and organized.

ACK N OWLED G EM ENTS
The Mediterranean LEK initiative was initially conceived by the inter- 1MED15_3.2_M2_337), and 85% cofunded by the European Regional Development Fund, which implemented the use of standard LEK protocols for marine protected areas and supported the writing of this publication. We warmly acknowledge the contribution of all the Mediterranean fishers, who shared with us their observations as well as all the esteemed researchers who adopted the LEK protocol, contributing to foster the use of this methodology at the Mediterranean level.

E TH I C A L S TATEM ENT
Data collection was confidential, as interviewers did not record any sensitive personal information about respondents. At the beginning