Diversity and patterns of marine non- native species in the archipelagos of Macaronesia

Aims: The present study is the first attempt to grasp the scale and richness of marine biological invasions in Macaronesia. We pioneered a comprehensive non- native species (NNS), inventory in the region to determine their diversity patterns and native distribution origins. NNS were defined here as the result of both introductions and range expansions. We also used statistical modelling to examine relationships among NNS richness, anthropogenic activities, demographic and geographical variables across Macaronesia.


| INTRODUC TI ON
Anthropogenic pressures cause significant impacts on global biodiversity and ecosystem structure and function (Costello et al., 2010;Mammides, 2020;Thuiller, 2007). Along with habitat degradation, the introduction of non-native species (NNS) is considered one of the greatest environmental and economic threats (Costello et al., 2010;Cuthbert et al., 2021;Diagne et al., 2021). In the last several decades, new estuarine and marine coastal NNS appear to be establishing worldwide at an increasing rate (Tsiamis et al., 2019). For example, NNS numbers increased considerably in European marine waters, reaching up to 824 NNS in 2018, making these waters the highest NNS host worldwide (Katsanevakis et al., 2014;Tsiamis et al., 2019). The vast preponderance of marine NNS established in Europe originates from the Western and Central Indo-Pacific, Temperate Northern Pacific, Tropical Atlantic and Temperate Northwest Atlantic (Tsiamis et al., 2018).
Biological invasions (the arrival, establishment and diffusion of a species) in marine communities occur through two processes, range expansions and introductions. Range expansions consist of dispersal by natural mechanisms into a region where the species did not formerly exist. Introductions consist of transportation by human activity (often across natural barriers) into a region where the species did not exist in historical time (although, in rare instances, the species may have existed in the region in geological time) (Carlton, 1987). Thus, the result of both range expansions and introductions is the arrival and potential establishment of non-native species (Carlton, 1987).
On a global scale, most marine invasions have resulted from the unintentional transfer of large numbers of animal and plant species in ballast water and hull fouling of commercial shipping (Bailey et al., 2020;Chainho et al., 2015;Ruiz et al., 1997). Hull fouling on recreational vessels also accounts for primary and secondary introductions (Ferrario et al., 2017;Marchini et al., 2015;Zabin et al., 2014).
Oceanic islands have long served as important models for biogeography, ecology, evolution and conservation biology (Darwin, 1859;Kueffer et al., 2014;MacArthur & Wilson, 1963;Wallace, 1880). Many oceanic islands possess high conservation status due to their endemic species richness (Sax & Gaines, 2008). However, islands have long been used for multiple purposes (i.e. farming, lighthouse stations, prisons, defence emplacements, tourism and more).
These activities contribute to the destruction of natural ecosystems and often lead to the introduction of NNS in both terrestrial and aquatic ecosystems (Veitch, 2001). Terrestrial introductions have been well-documented over the last 50 years on many island ecosystems and continue to be the focus of extensive work in invasion biology (e.g. Diamond, 1970;Elton, 1958;Gaston et al., 2008;Lloret et al., 2005;Rojas-Sandoval et al., 2020;Simberloff, 1995;Towns et al., 2012). The study of marine invasions on most of the world's islands only started much later, mainly in the last two decades, in Australia (Hewitt, 2002;Hewitt et al., 2004), New Zealand (reviewed by Inglis et al., 2006), Hawaii (Carlton & Eldredge, 2009, Guam (Paulay et al., 2002), Palau (Campbell et al., 2016), the Galápagos (Carlton et al., 2019) and in the Macaronesia region, particularly in the Azores and Madeira (Canning-Clode et al., 2013;Cardigos et al., 2006;Chainho et al., 2015;Micael et al., 2014).
Located in the Northeast Atlantic Ocean, the Macaronesia region includes the volcanic archipelagos of the Azores, Madeira, Canary Islands and Cabo Verde. These archipelagos are separated from the nearby mainland and other plausible natural species source regions, such as neighbouring archipelagos or shallow seamounts, by water depths exceeding 1300-1500 m (Freitas et al., 2019). Macaronesia's island systems are interconnected through ocean currents, with surface currents generally moving from the Azores to Madeira and Canary Islands (Morton et al., 1998;Santos et al., 1995). Moreover, Cabo Verde separates two main water masses: the southern boundary of the North Atlantic Subtropical Gyre (NASTG), which is here formed by the North Equatorial Current, and the northern edge of the North Atlantic Tropical Gyre (Pelegrí & Peña-Izquierdo, 2015).
In the marine realm, the Azores, Madeira and the Canary Islands are included in the same ecoregion, the Lusitanian province, dominated by rocky reefs. In contrast, Cabo Verde belongs to the West African Transition province (Spalding et al., 2007;Tuya & Haroun, 2009). The latter is under the influence of a more tropical climate, supporting, for example, hermatypic corals. Macaronesia province falls in the West Africa bioregion in IUCN Bioregions (Kelleher et al., 1995).
The Macaronesian Islands all have different degrees of isolation from the nearest continental coast. Such differences have been advocated to explain variations in composition and diversity of marine biota (Hawkins et al., 2000). Previous studies have highlighted a high degree of similarity in the marine flora's composition in each archipelago and the nearest continental (donor) coast. For example, the marine flora of the Azores has elements in common with the North Atlantic, the Western Mediterranean and the coasts of Eastern America (Frud'homme van Reine, 1988), while the Canary Islands shows a greater affinity with the Western Mediterranean and Western Atlantic (van den Hoek, 1987). Recent work proposed newly born biogeographical units excluding the Azores and Cabo Verde from the biogeographical concept of Macaronesia by applying extensive datasets from various marine taxonomic groups (Freitas et al., 2019).
As for marine biological invasions, the Azores was the first Macaronesian archipelago to be the subject of a detailed NNS inventory (Cardigos et al., 2006), followed by Madeira (Canning-Clode et al., 2013), both of which updated in Chainho et al. (2015). For the Canary Islands, a book published in 2003 listed marine species present in the archipelago, with a simple indication of each species indigenous or non-native statuses (Moro et al., 2003). For Cabo Verde, no marine NNS inventory has ever been produced. In recent years, because of ongoing monitoring surveys, several new NNS detections have been documented in the Azores, Madeira, the Canary Islands and a few in Cabo Verde (e.g. Afonso et al., 2013;Canning-Clode et al., 2013;Freitas et al., 2014;Freitas & Wirtz, 2018;García-Jiménez et al., 2008;Micael et al., 2014;Pajuelo et al., 2016;. However, no previous study has ever attempted to compile the current knowledge on marine NNS in all Macaronesia or relate diversity and distribution of NNS to historical or extent human activities.
In this context, the present study expands the current understanding of marine biological invasions' scale and diversity on insular marine ecosystems, using Macaronesia as a model system.
Specifically, we (1) conducted the first comprehensive NNS inventory of Macaronesia; (2) assessed NNS richness patterns across archipelagos and individual islands; (3) identified the possible native distribution of the documented NNS and (4) linked NNS numbers to anthropogenic, demographic and geographical variables.

| Study region
The Macaronesia region comprises four volcanic archipelagos-the Azores (nine populated islands), Madeira (two populated islands), the Canary Islands (seven populated islands) and Cabo Verde (nine populated islands) (Figure 1). The study region is located in the Atlantic Ocean between 15° and 40°N latitude, with distances from the European or African continents varying from 95 to 1600 km (Aranda et al., 2014). The geological ages of the islands range from 0.18 million years (MY) for Pico (Azores)  to 27 MY for Selvagem Grande (Madeira) (Aranda et al., 2014). Macaronesia's native flora and fauna reached the islands by long-range dispersal from adjacent continental areas (Whittaker & Fernández-Palacios, 2007) or, in some circumstances, from neighbouring archipelagos/ islands (Domingues et al., 2008). Finally, only populated islands in Macaronesia were considered for the present analysis.

| Literature search
In addition to our previous knowledge of published literature, a comprehensive literature search was conducted of scientific papers, books, book chapters, theses and reports. This search included literature published between 1880 and May 2020 in English, Portuguese and Spanish. Web of Science database, Scopus and Google Scholar were examined using the following relevant keywords (and/or): "alien", "invasive", "introduced", "non-indigenous species", "invasion", "non-native", "exotic", "Cabo Verde", "Azores", "Madeira", "Canary Islands" and "Macaronesia". To avoid any bias with medical sciences (i.e. cancer research) due to the use of the terms "invasive" and "invasion", the search was focused on the following research disciplines: ecology, biology, marine biology, fisheries, biodiversity, conservation, environmental sciences, oceanography and zoology. This effort was complemented by local experts, all of which are contributors to the present work. All references were carefully examined for marine NNS records in Macaronesia, and relevant subsequent citations were also analysed. In total, 200 references (Appendix S1) were validated and included in our study. For the present study, marine birds, marine mammals and vascular plants were not included in the search but would bear future work exploration. Finally, brackish and freshwater species were also excluded from the present analysis.

| NNS selection and attributes
For the present work, a species was considered non-native (NNS), and therefore validated (included in the analyses-Appendix S1-"1" included species), when as many as possible of the following criteria were met: (i) reference of its biogeographical status in the literature; (ii) indication that a species found either only or mainly in ports and marinas, as these areas have higher propagule pressure and are considered NNS hotspots (Canning-Clode et al., 2013;Seebens et al., 2013); (iii) expert opinion and/or reference in known marine biological invasions databases (e.g. NEMESIS [Fofonoff et al., 2018], AquaNIS [AquaNIS. Editorial Board, 2015]); (iv) species that underwent tropicalization processes (i.e. shifts in range distribution induced by climate change [Canning-Clode & Carlton, 2017]) and (v) current NNS population status. The use of each criterion is described in Appendix S1-Column 'Criteria' for each species validated.
Relative to criterion (i), we are aware that there is a long history of misconstruing native species as introduced species (i.e. pseudoindigenous, Carlton, 2009;Carlton & Eldredge, 2009). Thus, we typically relied on the additional evidence outlined above and below for validating NNS. Respectively to criterion (ii), we recognize that native species are also found in port facilities, and intense studies of port systems might reveal native species that are not yet reported elsewhere. However, we use multiple criteria and do not use 'only in ports' as the sole criterion by which to identify a species as NNS. Regarding criterion (v), a species was removed from the analysis when the record in Macaronesia was based on single or few specimens (less than five) collected at one or few locations (less than three). Examples of removed species include the nudibranch Antiopella cristata (Delle Chiaje, 1841), the hydroid Tubularia indivisa Linnaeus, 1758 and the phoronid Phoronis hippocrepia Wright, 1856 (see Appendix S1 for a complete listing). When species with native or non-native status in the literature were doubted (e.g. the bryozoan Aetea spp., the tunicate Botryllus schlosseri (Pallas, 1766) and the hydroid Corydendrium parasiticum (Linnaeus, 1767)) further assessment based on historical, systematic, biogeographic and other criteria were conducted (Chapman & Carlton, 1991. We followed Carlton (1996) in defining species that were not demonstrably native or non-native as cryptogenic, and therefore those were not included in our analysis.
We are sensitive to the concern that records of apparently new additions to a biota may be previously overlooked species due to a lack of habitat-specific studies (such as in deeper waters) or to the earlier absence of taxonomic specialists. In these cases, we have attempted to rely on evidence of the previous absence of the species in F I G U R E 1 Location of Macaronesia and its archipelagos (Azores, Madeira, Canary Islands and Cabo Verde) question as best as possible. Critically, we emphasize that we did not attempt at this time to re-evaluate or to re-categorize many species that are treated as 'native' in the Macaronesia literature that may be non-native or cryptogenic. Many species introduced before modern surveys began and long regarded in the literature as native but now re-interpreted as non-native form a substantial fraction of current inventories of NNS in other archipelagos, such as the Hawaiian Islands (Carlton & Eldredge, 2009 and the Galapagos Islands (Carlton et al., 2019). Future work in Macaronesia will address such re-evaluations.
The year of the first record for each NNS and cryptogenic species was retrieved from the literature. The species where the first observation date was not mentioned in the bibliography were removed from the analysis but kept in the overall list. When only time ranges were provided (e.g. '1999-2003'), the first year of that time interval was taken as the earliest possible date. Also, species whose records only included reference to the archipelago and not the island (i.e. Azores or Cabo Verde) were excluded from the model but included in the final list. Taxonomic groups were categorized as NNS, and attention has been taken to standardize nomenclature.
Taxonomic references were updated according to the Integrated Taxonomic Information System (www.itis.gov), the World Register of Marine Species (Worms Editorial Board, 2020), and in the case of algae, according to AlgaeBase (Guiry & Guiry, 2020).
For each NNS, the likely native distribution range (origin) was assigned using the 18 large-scale IUCN marine bioregions as defined by Kelleher et al. (1995) and later modified by Hewitt and Campbell (2010)
ine.cv (Cabo Verde)), provided geographical and demographic variables such as island area (km 2 ), island human population (number of individuals) and human population density (number of individuals/km 2 ). Variables related to marine traffic for each island were used, including the number of ports, number of marinas, the sum of ports and marinas, the number of major marinas (defined as with more than 120 berths), total port area (km 2 ), total port perimeter (km), total marinas area (km 2 ), the sum of berths in all the marinas across islands and the sum of ships departures per month in each island. The number of ports and marinas on each island and the num-

| Data analysis
Presence/absence NNS matrices were created to quantify biological invasions in the Macaronesia region. A standard exploratory data analysis was conducted, comparing NNS richness and taxonomic group per archipelago. Non-metric multidimensional scaling (MDS) was applied to visualize the similarity between islands, and multivariate analysis of similarity (ANOSIM) was used to test community composition between archipelagos (e.g. Jaspers et al., 2020).
The available covariates were used to model the number of NNS per island. Given the nature of the response, the variable is a count (i.e. number of NNS), so we considered a Generalized Linear Model (GLM) (Zuur et al., 2009). We deemed both Poisson and Negative Binomial (NB) responses. After an exploratory examination of the relative merits of both distributions, the NB was chosen. The variance was higher than the mean, and hence not surprisingly, the NB led to far more parsimonious models than the Poisson given Akaike Information Criterion (AIC). We considered a stepwise procedure for model selection based on AIC (e.g. Burnham et al., 2011). Given the small number of observations (i.e. 27 islands), we started from the simplest model and increased its complexity incrementally until adding new covariates no longer increased parsimony. All covariate pairs were checked for correlation, which was generally large across pairs of variables. Hence, we avoided considering highly correlated covariates in the same model, always selecting the one from each pair under consideration to lead to a more parsimonious interpretation.
We assessed the fit of the final model used for inference based on visual inspection of residual diagnostic plots. All the analyses were implemented in R (R Core Team, 2019). Multivariate analysis was implemented via the 'vegan' R package (Oksanen et al., 2012), while the 'mass' package (Ripley et al., 2019) was used for GLM application.

| RE SULTS
In total, 144 NNS were detected and considered for the whole Macaronesia region (Figure 1; Appendix S1). The Canary Islands recorded the highest number (76 NNS), followed by the Azores (66) for all the archipelagos, except for Madeira and Cabo Verde that only increased their NNS numbers during the 1990s.
Our search detected 46 additional species that did not fully meet our criteria and were not included in our analysis (complete detailed list in Appendix S1 including records not considered for analysis purposes (i.e. cryptogenic species and unestablished NNS). In some cases, registered NNS were deemed to be native or cryptogenic in different archipelagos of Macaronesia. In this context, the Canary Islands registered the highest number of cryptogenic species (nine), followed by Cabo Verde (six), Madeira (five) and the Azores, with only one species. Also, Cabo Verde and the Canary Islands were the archipelagos with the highest number of native species, 16 and five respectively. Madeira registered only one native species and none in the Azores (Appendix S1).
For the whole Macaronesia system, macroalgae were the most represented non-native taxonomic group (31), followed by tunicates (28), fishes (Vertebrates) (26) and bryozoans (14) (Figure 3). In contrast, Ctenophora was the least reported taxonomic group with only one NNS record. The Azores had a similar trend to the overall Macaronesia pattern, with macroalgae (22), tunicates (11), bryozoans and arthropods (nine each) as the most observed taxonomic groups.
Fishes were the exception of the overall Macaronesia pattern, with only one species registered in the Azores. The Madeira archipelago had a higher number of tunicates (13), followed by bryozoans (12), fishes (nine) and macroalgae (eight) (Figure 3). For the Canary Islands, macroalgae and fishes were the most significant taxonomic groups with 20 and 18 NNS, followed by cnidarians (10) and tunicates (eight).
Finally, the latter group was the most important taxa in Cabo Verde (11), followed by bryozoans (two), and macroalgae, fishes, sponges, cnidarians and arthropods with one NNS each (Figure 3). For statistical modelling, multicollinearity was quite extreme for some of the available variables used to explain the number of NNS per island, particularly in covariates related to marine traffic ( Figure 1 in Appendix S4). Multicollinearity was also high for the different distance variables associated with isolation. For the model implementation, only the minimum distance to the mainland for each island was applied. We considered this to be the best strategy to incorporate the isolation information on the distances while resolving the extreme multicollinearity challenge. The stepwise selection was used to remove marine traffic correlated covariates. The best model based on the stepwise procedure was an NB model where NNS richness is described as a function of (1) minimum distance to the mainland, (2) each archipelago, (3) total marinas area (km 2 ) and (4) the total number of ports and marinas (Table 1 and Table 1  Vicente (Cabo Verde) were the islands with more underestimated NNS numbers (Figure 7). For more details in predicted values for each island, see Table 2 in Appendix S4.

| DISCUSS ION
This study expands the current understanding of the scale and di-   (Kelleher et al., 1995) and later modified by Hewitt and Campbell (2010). Note that some species have more than one native origin. Bioregions codes as follows: 1-Antarctica (Ant); 2-Arctic (Arc); 3-Mediterranean including the Black and Azov Sea (Med); 4-North West Atlantic (NWA); 5-North East Atlantic (NEA); 6 -Baltic (B); 7-Wider Caribbean Sea (WCS); 8-West Africa (WA); 9-South Atlantic (SA); 10-Central Indian Ocean (CIO); 11-Arabian Seas (AS); 12-East Africa (EA); 13-East Asian Seas (EAS); 14a&b-South Pacific & Hawaii (SP); 15-North East Pacific (NEP); 16-North West Pacific (NWP); 17-Southeast Pacific (SEP); 18 -Australia and New Zealand (Aus) (see Appendix S2 for further details) Most NNS present in Macaronesia were native to West Africa, Australia, New Zealand and the wider Caribbean Sea and North West Pacific bioregions, followed by the North West Atlantic, and the Mediterranean. The Azores differs from the most with slighter Australia affinities and the Canary Islands with more West Africa signature. Finally, NB modelling suggested that non-native richness patterns across Macaronesia were strongly affected by (i) minimum distance to the mainland, (ii) each archipelago, (iii) total marinas area (km 2 ) and (iv) the total number of ports and marinas.
By using statistical models, we simplified reality, which assists its interpretation and facilitates exploiting results as a valuable tool to construct and explore different hypothetical scenarios (e.g. Walsh & Brodziak, 2015). In this context, our study predicted the number of NNS present in each Macaronesia Island as a function of geographical variables and coastal development elements. The sample size available was small for modelling routines (27 islands), but we included multiple available covariates. Hence, model fitting and model selection was challenging, and minor changes to the species list tended to lead to different variables being included in the best model, mainly including marine traffic facilities (ports and marinas) and island population ( Table 1 in Appendix S4). With such a small sample size, removing observations is far from easy and probably not recommended. The variable total island area was not significant in the present model, but larger islands had a higher NNS number (except Cabo Verde). Usually, larger islands are associated with more human activities and development, resulting in an enhanced anthropogenic disturbance (higher propagule pressure), and consequently, more NNS introduction events (Rojas-Sandoval et al., 2020).  (Micael et al., 2014) with an ostensible cosmopolitan distribution (Andreakis et al., 2007;Chualain et al., 2004).

| Overall Macaronesia context
Macaronesia's introduction of this macroalgae species was likely via ship hulls or rafting from other invaded regions (Cardigos et al., 2006). The spaghetti bryozoan A. verticillata registered recent introduction events in all Macaronesia archipelagos except in Cabo Verde (Minchin, 2012 and Appendix S1), where it was first collected in 1904 (Waters, 1918).  (Marchini et al., 2015). For the bryozoan S. errata, we considered earlier reports of Schizoporella unicornis (Johnston in Wood, 1844), a colder water European species, from Macaronesia, to very likely represent S. errata, with which it was long confused (Tompsett et al., 2009; see also Ryland et al., 2014). The bryozoan S. errata is a warmer water European species, which we regard as NNS in Macaronesia, where it has most likely been transported by ship fouling (Carlton & Eldredge, 2015). The barnacle B. trigonus of Pacific origin has spread over the Atlantic before the 1900s, most likely on ship hulls . This species can be found on both sides of the Atlantic and throughout the Mediterranean (Fofonoff et al., 2018). It is a common element in fouling communities, and likely shipping and aquaculture were the introduction vectors in Macaronesia (Cardigos et al., 2006;Chainho et al., 2015;Fofonoff et al., 2018 (Garcia, 2017). The Macaronesian archipelagos and other Atlantic islands provided 'port of call' facilities (Garcia, 2017). The triangular commercial sailing route in the late 1800s between Europe, West Africa and the Caribbean (Crosby, 1986) and the British steamer routes and coaling stations (Mack, 2003) could have enhanced the connectivity amongst areas never joined before. Evidence suggests that the first NNS was detected in the late 19th and beginning of the 20th centuries . Still, it cannot be excluded that some species, considered now as native species, might be earlier introductions from the 15th and 16th centuries. For example,
The similarity of NNS and the shared NNS among Macaronesian archipelagos could be related to several factors associated with isolation, for example, distance to the mainland and/or amongst islands and/or marine traffic connectivity and intensity. The present study verified the highest similarity in NNS diversity between Madeira and Canary Islands, with 30 shared NNS. These archipelagos are the closest ones in the whole Macaronesia province. They have long been connected as stopovers for yachts crossing the Atlantic from the East (Parrain, 2011) and cruise ship routes (Sousa, 2000). The Azores and the Canary Islands had 27 unique shared NNS. In this latter example, proximity may not be the best explanation for these similarities, which can be better explained by the high number of studies focusing on macroalgae diversity and taxonomy, which have identified 12 macroalgae species that are present in both archipelagos (Appendix S1).
Cabo Verde, the southernmost archipelago of Macaronesia, was particularly differentiated from the remaining archipelagos, both in NNS composition similarity and in shared NNS. Interestingly, Cabo Verde has more shared NNS (10) with Madeira than with the Canary Islands (nine), its closest neighbour, which may be partly explained by the fact that Cabo Verde is a former Portuguese colony that had frequent connections with other Portuguese territories, including Madeira archipelago .
Each archipelago had some differences related to NNS origin.
The Azores are more influenced by Australia and the Pacific, Madeira According to some studies, the most recent introductions often F I G U R E 7 Non-native species (NNS) detected in the present study (black colour) and by the output results predicted by the selected Negative Binomial (NB) model (grey colour) for each island of the four Macaronesian archipelagos. When the models' prediction (grey colour) is not visible, the observed value (black colour) overlaps the predicted value. Predictions close to observed values might, therefore, not be visible result from secondary introductions (Chainho et al., 2015;Martínez-Laiz et al., 2020;Souto et al., 2018). Finally, the distance to neighbouring continents and islands that initially affected native species colonization by dispersion (Domingues et al., 2008;Whittaker & Fernández-Palacios, 2007)

| NNS patterns in Azores
The Azores registered 66 NNS, being the second archipelago with higher NNS introductions in Macaronesia. In this Portuguese archipelago, macroalgae species (22), tunicates (11), bryozoans and arthropods (9) were the most relevant non-native taxa, similar to the overall pattern of the whole Macaronesia. Studies in marine ecology in the Azores started earlier than in other Macaronesian archipelagos, and there was some interest in marine species introductions as early as in the 1970s (e.g. Monniot, 1971;Morton & Britton, 2000;Tittley & Neto, 1994). In 2006, Cardigos et al. (2006 produced the first NNS inventory for the Azores based on scientific publications, reports and personal data. The work of Cardigos et al. (2006) increased the interest in the study of marine biological invasions in the Azores, and several other papers have been recently published (e.g. Micael et al., 2017;Vaz-Pinto et al., 2014) and have increased the knowledge in marine species introductions in the archipelago.
In 2014, a review by Micael et al. (2014) confirmed that the Azores have more macroalgae introductions than other parts of the globe.
Marine shipping (mainly through ballast water and hull fouling) is considered the primary introduction vector of NNS into the Azores (Cardigos et al., 2006;Micael et al., 2014). Moreover, the Azores have been selected as a critical destination for transatlantic recreational boating over the years, increasing the likelihood of NNS introductions (Cardigos et al., 2006). Other vectors have also been listed as relevant in facilitating the introduction or spread of NNS in the Azores, including aquarium trade and boating-related scuba diving activities (Cardigos et al., 2006;Parretti et al., 2020). Some NNS detections were a consequence of monitoring programmes that

| NNS patterns in Madeira
Our search confirmed the Madeira archipelago as the third Macaronesian archipelago in NNS numbers. Madeira has a few research institutions partially working on marine sciences. With a few sporadic records detected during the 1990s (e.g. Wirtz, 1995Wirtz, , 1998 ductions in Madeira were also related to tropicalization processes (Ribeiro et al., 2019;Schäfer et al., 2019). Finally, to a lesser extent, other vectors have been suggested to be facilitating NNS introductions in Madeira, including aquaculture (Alves & Alves, 2002) and, more recently, transatlantic rafting (Wirtz & Zilberg, 2019).

| NNS patterns in the Canary Islands
The highest number of NNS was observed in the Canary Islands, and several factors could account for this observation. First, the most frequent taxa detected in the Canary Islands were macroalgae and fish. The biogeographical position of the Canary Islands has been recognized as the major element for the richness of its marine biota (Haroun & Herrera, 2001), with a high interest in phycology, where several publications were produced over the years (e.g. Afonso-Carrillo et al., 2007;Gil-Rodríguez & Afonso-Carrillo, 1980;Haroun & Herrera, 2001;Sangil et al., 2012). Therefore, it is no surprise that macroalgae are amongst the most detected NNS taxa in this archipelago. Recently, the number of non-native fishes detected in the Canary Islands increased considerably (e.g. Falcón et al., 2015;Pajuelo et al., 2016). Our study reports 18 established non-native fish species (see Methods section for more details) in Canarian waters, most of these associated with oil platforms (e.g. Brito et al., 2011;Falcón et al., 2015;Pajuelo et al., 2016;Triay-Portella et al., 2015).  (Brito et al., 2011;González et al., 2017).
Additionally, other human-induced vectors may be responsible for some of these introductions in the Canary Islands, including aquarium trade (Falcón et al., 2015) and aquaculture (Toledo-Guedes et al., 2014). Besides, range expansions due to anthropogenic-induced climate change could have influenced these numbers (Brito et al., 2005).
One other element that may affect NNS distribution per country is the variability in the monitoring and reporting effort (Katsanevakis et al., 2013). In this context, several governmental institutions and Universities in the Canary Islands have been conducting studies related to marine biological invasions. Moreover, the Canary Islands have the highest human population density in Macaronesia, with an estimated 2 million residents and around 10 million tourists every year (Garín-Muñoz, 2006), which exposes these islands to elevated anthropogenic pressure. The Canary Islands hold more ports and marinas and a superior port area among the four Macaronesian archipelagos. This suggests higher propagule pressure (e.g. number of viable NNS individuals, the number of discrete introduction events, their frequency and duration) which is recognized as the primary determinant of NNS invasion success (Ojaveer et al., 2014). Besides, this extended port area, dominated by several artificial structures like piers, pontoons, seawalls and buoys, provides transport hubs or 'stepping stones' for potentially newly established species (Bulleri & Chapman, 2010;Pinochet et al., 2020;Ruiz et al., 2009). Some studies point out that artificial substrates expedite the NNS colonization process compared to natural substrates (Lambert & Lambert, 2003;Pinochet et al., 2020;Tyrrell & Byers, 2007).

| NNS patterns in Cabo Verde
Cabo Verde was the archipelago with the fewest NNS detections in our search. In addition to marine traffic (e.g. Monteiro, 2012) (Freestone et al., 2013;Gestoso et al., 2017Gestoso et al., , 2018.
However, likely, the number of reported NNS in Cabo Verde is severely influenced by reduced sampling effort in the archipelago.

| Conclusions
Although not detected as a significant component in our model, study effort could also play an essential part in NNS detection.
Moreover, NNS findings are usually linked with monitoring and reporting (Chainho et al., 2015;Katsanevakis et al., 2013). In this context, future research in Macaronesia should focus on standardized NNS monitoring surveys using standard protocols.
Coastal development, that is, ports and marina infrastructures, favours species establishment and potential dispersal to adjacent areas (Afonso et al., 2020). Marine traffic played and still plays a vital role in species arrival and dispersal. The 144 NNS listed here will likely increase the awareness regarding marine NNS in the whole Macaronesia region serving as a trigger for future works and implementing and enforcing regulations addressing the introduction of marine NNS in oceanic islands. Given that the majority of marine NNS recently recognized in the Hawaiian Islands (Carlton & Eldredge, 2009 and Galapagos Islands (Carlton et al., 2019) are those that were previously listed solely as native species and did not appear in any lists of NNS for these archipelagos, we assumed that the number of NNS that we recognized here is likely to be a fraction of the actual NNS diversity in Macaronesia. The overall number of NNS in Macaronesia that we present here will probably never be definitive as new species are constantly arriving and being detected (Álvarez-Canali et al., 2021), or more detailed studies are conducted (Ramalhosa et al., 2021).
In the present paper, NNS reported numbers are dependent on each archipelago, and strongly affected by total marinas area, the total number of ports and marinas, and mean distance to the closest continental landmass. This suggests that more developed islands with higher marine traffic intensity and more prominent port infrastructure host more NNS.
Finally, millions of years of physical isolation have favoured the evolution of unique species and habitats in oceanic islands, which can be quickly exposed to an increasing number of NNS threatening native species and even driving some species to extinction (Micael et al., 2014). Thus, the pressure on the islands' endemic biota is likely greater than the one reflected by the numbers at this moment presented. Nevertheless, using Macaronesia as a study model, the present work represents a pioneer effort to characterize and better understand marine invasions in the northeast Atlantic insular ecosystems. This effort will undoubtedly contribute and serve as a baseline for future ecological, experimental and management studies.

CO N FLI C T O F I NTE R E S T
The authors declare there is no conflict of interest.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/ddi.13465.

DATA AVA I L A B I L I T Y S TAT E M E N T
All newly generated data used in this study are available as supplementary electronic material (Appendices S1-S4).