The genus Carpobrotus N.E.Br. comprises between 12 and 25 species, most of which are native to South Africa. Some Carpobrotus species are considered among the most damaging invasive species in coastal dune systems worldwide. In their introduced areas, these species represent a serious threat to native species and significantly impact soil conditions and geochemical processes. Despite being well studied, the taxonomy of Carpobrotus remains problematic, as the genus comprises a complex of species that hybridize easily and are difficult to distinguish from each other. To explore the population genetic structure of invasive Carpobrotus species (i.e., C. acinaciformis and C. edulis) across a significant part of their native and non-native ranges, we sampled 40 populations across Argentina, Italy, New Zealand, Portugal, South Africa, Spain, and the USA. We developed taxon-specific microsatellite markers using a Next Generation Sequencing approach to analyze the population genetic structure and incidence of hybridization in native and non-native regions. We identified three genetically distinct clusters, which are present in both the native and non-native regions. Based on a set of selected morphological characteristics, we found no clear features to identify taxa morphologically. Our results suggest that the most probable sources of global introductions of Carpobrotus species are the Western Cape region of South Africa and the coastline of California. We suggest that management actions targeting Carpobrotus invasions globally should focus on preventing additional introductions from the east coast of South Africa, and on searching for prospective biocontrol agents in the Western Cape region of South Africa.

Biological invasions, genetic diversity, genetic structure, hybridization, introduction history, invasive alien plant, microsatellite markers, taxonomic uncertainty

Coastal habitats such as coastal dunes, sea cliffs, and coastal prairies are exposed to a variety of extreme environmental conditions, including high salinity, low soil moisture, soil nutrient deficiencies, and intense wind and solar irradiance (Maun 2009). These conditions result in a high degree of specialization among species that naturally occur in these habitats (Mayoral et al. 2021). As such, coastal areas often host rare and endemic communities of high conservation value (Acosta et al. 2009). But coastal areas are also among the most endangered habitats (Defeo et al. 2009) and several anthropogenic drivers threaten their conservation, including biological invasions, climate change, habitat degradation, and urbanization (Carboni et al. 2009; Dawson et al. 2017). Invasive plants are considered to be one of the main threats to the conservation of the biodiversity and ecosystem functioning of coastal areas across the world (Millennium Ecosystem Assessment 2005).

The succulent genus Carpobrotus N.E.Br. (family Aizoaceae) comprises between 12 and 25 species and lower-rank taxa, most of them native to South Africa (Hartmann 2002). Several of these species are considered to be among the most widespread and damaging invasive plants in coastal areas globally (Campoy et al. 2018). Carpobrotus taxa have been introduced to coastal areas accross the world for ornamental purposes and for soil and dune stabilization. For example, they have been present in European gardens since the late 17^th century (Preston and Sell 1988) and, in California, they have been used for soil stabilization since the early 20^th century (Albert et al. 1997). Carpobrotus species have invaded millions of hectares of coastal areas worldwide, including in Argentina, Australia, California, Chile, New Zealand, and Southern and Western Europe (Campoy et al. 2018), impacting biodiversity and native species community structure and ecosystem functioning in multiple ways. For example, they compete with native plants for space, nutrients and water, reducing their growth, survival, and reproduction (D’Antonio and Mahall 1991; Molinari et al. 2007; Novoa and González 2014). They are also considered ecosystem engineers (Cuddington et al. 2011) since they can cause substantial and irreversible changes to invaded soils (Novoa et al. 2014). In particular, dense patches of invasive Carpobrotus produce and accumulate large amounts of litter (Fenollosa et al. 2016), which increases soil water holding capacity and, during its decomposition, decreases soil pH, and increases soil nutrient contents (Novoa et al. 2012, 2014). These changes ‘soften’ the extreme environmental conditions typical of coastal areas and facilitate the establishment and growth of opportunistic weeds while replacing native coastal vegetation (Novoa et al. 2012, 2013). Invasive Carpobrotus also alters the diversity, composition and functioning of soil microbial (Lechuga-Lago et al. 2017; Novoa et al. 2020) and invertebrate communities (Rodríguez et al. 2020; Gutiérrez 2021) and disrupts native pollination (Jakobsson et al. 2008) and herbivory networks (Rodríguez et al. 2019, 2021).

To gain insight into the invasiveness and impact of non-native species, as well as to develop or improve management actions it is important to know the taxonomic identity and the introduction history of the target invasive species (Pyšek et al. 2013). However, the taxonomy and biogeography of Carpobrotus spp. have long been a subject of debate (Campoy et al. 2018). Most of the taxa are native to South Africa, but five are native to Australia, and one species (C. chilensis) may be native to the Americas. Carpobrotus spp. have been described in several floras worldwide (Harvey and Sonder 1861; Blake 1969; Bolus Herbarium Collection 2015; Preston and Sell 1988; Gonçalves 1991; Wisura and Glen 1993), but these lists do not use the same traits to delineate species. The main diagnostic characters used to differentiate species are flower color and shape of the leaf section. However, there are doubts over the validity of these traits for identifying Carpobrotus species (Campoy et al. 2018). Thus, the information given in these documents cannot be easily synchronized or compared (Hartmann 2002). Moreover, due to their succulence, Carpobrotus spp. are difficult to curate, and therefore are poorly represented in herbarium collections (Walters et al. 2011). In fact, in several cases, the species names are based on lectotypes selected from illustrations, e.g., by Dillenius (1732). As a result, the taxonomy of the genus remains problematic.

Two Carpobrotus species are currently considered to be invasive: C. edulis (L.) N.E.Br., and C. acinaciformis (L.) L.Bolus (Campoy et al. 2018). Carpobrotus edulis is the most popular and widely introduced species in the genus. It is native to South Africa and considered one of the worst invasive plants of coastal areas and one of the most thoroughly studied invasive species worldwide (Pyšek et al. 2008; Campoy et al. 2018). It has been reported to hybridize with other Carpobrotus species in its native and invasive ranges (hybrids have been documented in the Americas, Australia, Europe, and South Africa; e.g., Campoy et al. 2018). Hybrids between C. edulis and species from other genera (e.g., Sarcozona) have also been reported outside South Africa (e.g., Heenan and Sykes 2010). Carpobrotus acinaciformis is generally considered to be native to South Africa, although it has also been suggested that it may be a hybrid between C. edulis and other South African or Australian congeners (Schierenbeck et al. 2005). Carpobrotus edulis and C. acinaciformis have a long history of human use in South Africa, and therefore, their natural limits and identities may also be conflated (Malan and Notten 2006).

Carpobrotus chilensis also provides a good example of the taxonomic and biogeographic uncertainties that plague the genus. Some authors consider this species to be native to California and Chile (Brown 1928), while others regard it as native to Argentina and Chile (Hartmann 2002; Zuloaga and Belgrano 2017; US National Plant Germplasm System 2022) and still others suggest it is “probably native to South Africa” (https://ucjeps.berkeley.edu/eflora/eflora_display.php?tid=77164). An extensive review of herbarium and historical records carried out to identify the origin of this species was inconclusive (Bicknell and Mackey 1988). Some authors have even considered it to be a hybrid swarm of five South African species (i.e., C. deliciosus, C. dimidiatus, C. edulis, C. mellei, and C. muirii) (Bicknell and Mackey 1998). Hence, the origins and taxonomic classification of this taxon are speculative at best. In California, C. chilensis has been reported to hybridize with the South African C. edulis (Gallagher et al. 1997; Albert et al. 1997; Vilà and D’Antonio 1998) with extensive directional backcrossing and potential loss of pure C. chilensis types (Vilà et al. 1998; Schierenbeck et al. 2005). Overall, the genus Carpobrotus is often considered to be a complex of species that easily hybridizes and are difficult to distinguish (Traveset et al. 2008). This taxonomic uncertainty is further complicated by the clonal growth typical of the genus, which stabilizes hybrid genotypes (Ellstrand and Schierenbeck 2000).

Here, we aim to shed light on the relatedness and introduction history of invasive Carpobrotus spp. around the world. With this overarching aim, we (1) sampled invasive Carpobrotus species in coastal areas across many of their presumed native and invaded ranges and (2) developed and used a set of genus-specific microsatellite markers to assess and compare the genetic diversity and structure among these populations. Moreover, aiming to help managers and other stakeholders with the identification of invasive Carpobrotus species in the field, we (3) compared the morphological characteristics of the Carpobrotus taxa assigned to distinct genetic clusters.

Our results confirm the complex identification, biosystematics and biogeography of the invasive Carpobrotus spp. The west coast of South Africa, and possibly California, were identified as the most likely sources of invasive populations worldwide.

The Bayesian assignment analysis grouped the sampled populations into three genetic clusters (clusters A, B and C; Fig. 3). In South Africa, the native distribution area of most Carpobrotus spp. (Germishuizen and Meyer 2003), most sampled individuals were assigned to clusters A (in the Western Cape province) and C (in the Eastern Cape and Kwazulu-Natal provinces). Four Carpobrotus species (including their described lower taxa) are considered native to the Western Cape province: C. acinaciformis, C. edulis, C. muirii, and C. quadrifidus (Smith et al. 1998). Individuals assigned to cluster A could therefore correspond to one or several of these species, or to hybrids between them. On the other hand, two species occur naturally in the Eastern Cape and KwaZulu-Natal provinces: C. deliciosus and C. dimidiatus (Smith et al. 1998). Therefore, cluster C likely corresponds to individuals of one or both of these two species or hybrids between them.

Only one South African population, consisting of a single genotype, was assigned to cluster B (shared by some populations from southern Europe and California; Table 1). This population is located in Mdumbi, a remote area that attracts tourists from all over the world due to the presence of various ecotourism establishments and surfing lodges (Hitchcock 2014). Cluster B was predominantly found in California, although two populations from Spain and one in the Azores were also from this cluster. These results suggest that populations assigned to cluster B might have originated from South Africa decades ago, introduced to California directly from South Africa or secondarily via Spain (i.e., a country with an extensive history of trade with the California coast; Engstrand 1997), and hybridized extensively (Vilà et al. 1998; Schierenbeck et al. 2005) with C. chilensis, a species of unknown origin that mainly occurs in the Pacific coasts of the American continent (Campoy et al. 2018). The abrupt appearance of C. chilensis pollen within a 900-year-old record for the central California coast in the early 1800s suggests introduction with early Spanish settlement or visitation including extensive migration of people from Portugal from the Azores to coastal California in the 1800s (Williams 1982; Bicknell and Mackey 1988). Individuals representing cluster B may then have been introduced to the coast of Spain and/or the Mdumbi region after hybridization had occurred with the later introduced C. edulis. It is also possible that the C. chilensis plants in California arose from an early introduction from an unknown source where the species no longer exists and then pure C. chilensis in California has largely disappeared through hybridization.

Outside South Africa and California, all sampled populations were assigned to either cluster A or B, or were identified as admixed. These findings suggest that the Western Cape province of South Africa and coastal California may have served as the sources for many introduced Carpobrotus populations in the rest of the world. This is not surprising, given that Carpobrotus species have been widely introduced as ornamental plants (Campoy et al. 2018), and both regions have been prominent hubs of the ornamental horticulture industry for centuries (University of California 1999). More specifically, all individuals sampled in New Zealand were assigned to cluster A, suggesting a South African origin of Carpobrotus invasions in this country. In Spain and Azores, most populations were assigned to clusters A and B, suggesting multiple introductions from South Africa and the Americas.

The Italian and Argentinian populations included in our analyses were not clearly assigned to particular genetic clusters, suggesting that genetically distinct groups or species of Carpobrotus were introduced to these areas from different sources, leading to extensive admixture (Suehs et al. 2004). Accordingly, hybridization has been repeatedly suggested to play an important role in the invasiveness of Carpobrotus species (Campoy et al. 2018), with hybrids having higher survival and faster growth rates than parental taxa (Vilà and D’Antonio 1998). Moreover, our results show that hybridization is also common in South Africa, the native range of most species in the genus. The implications of hybridization for the invasion of Carpobrotus are poorly understood and deserve further research attention.

Overall, our results indicate that there have been multiple introductions of Carpobrotus species from different sources globally. Typically, multiple introductions increase the genetic diversity and probability of success of invasive species (Genton et al. 2005; Walls 2010). However, we found extremely low levels of genetic diversity in all studied populations. The reason for this can probably be attributed to the capacity for self-fertilization (Vilà et al. 1998) and the clonal nature of Carpobrotus species, which facilitates vegetative reproduction without genetic recombination (Campoy et al. 2018) which typically results in low genetic diversity (e.g., Hollingsworth and Bailey 2000). Accordingly, we observed low inbreeding and high clonality levels in all sampled populations. Moreover, clonality has been suggested to allow Carpobrotus species, and alien plants in general, to effectively establish and colonize new areas (Roiloa et al. 2010). These observations also explain the high number of monomorphic loci we identified during genetic marker development and testing.

Accurate identification of invasive Carpobrotus species or hybrid combinations could improve risk assessment and guide early detection and rapid response management actions (Guisan and Thuiller 2005). For example, in California, some managers do not want to remove what seems to be C. chilensis because they do not know whether or not it is native and it appears to coexist with native species and can be helpful in dune stabilization (D’Antonio, personal observation). Also, “taxonomic identity” should be specified in any risk assessment/analysis scheme (e.g., IPPC, ISPM 2, Framework for pest risk analysis) and local management plans for the removal of species. Similarly, Species Distribution Models (SDMs) used to guide early detection and rapid response actions typically use distributional data of the target species, coupled with characteristics of the current and potentially suitable areas (e.g., climate, land-use type) (Guisan and Thuiller 2005). Using Ecological Niche Models, Thuiller et al. (2005) identified areas of high suitability for invasive Carpobrotus species in Australia, central east Africa, Chile, Europe and the USA. However, it is conceivable that the geographic extent of such predictions depends on the occurrence records of the Carpobrotus species and/or their hybrids used to calibrate these models. Accurate identification and knowledge of the introduction history of invasive Carpobrotus spp. are also critical for reducing the negative impacts of their current invasions. The most common methods used to control Carpobrotus invasions include mechanical and chemical methods (Ruffino et al. 2015). However, these methods require large amounts of funding and capacity, follow-ups and restoration efforts, and have not been successful at reducing Carpobrotus invasions at large geographic scales. The integration of biological control into the management of invasive Carpobrotus species could reduce management costs significantly and increase management success (Campoy et al. 2018). For example, the South African soft scale Pulvinariella mesembryanthemi (Vallot, 1829) is a specialist herbivore of Carpobrotus spp. that was accidentally introduced into California, causing considerable damage to invasive populations of C. edulis (Washburn and Frankie 1985) where it also became a pest of the presumed native C. chilensis (Schmalzer and Hinkel 1987). Subsequently, predators and parasites were released from South Africa to control the scale (Tassan et al. 1982). Pulvinariella mesembryanthemi is still a promising potential biological control agent outside of California (Vieites-Blanco et al. 2019; Núñez-González et al. 2021). But the effectiveness of P. mesembryanthemi is likely to depend on the taxonomic identity and source region of the target species (Pyšek et al. 2013; Le Roux 2021).

However, identifying invasive Carpobrotus species is challenging. Several diagnostic morphological characters have been proposed to differentiate between species, with petal colour being the most popular one (Preston and Sell 1988; Wisura and Glen 1993), but doubts have been expressed on the validity of all proposed characters as taxonomic markers (Campoy et al. 2018). Our results show no clear pattern regarding the association of morphological traits with the three genetic clusters we identified. We only collected morphological data from 10 individuals per population in the field, and each population was located in a different coastal habitat (e.g., disturbed areas or dunes). The different conditions to which Carpobrotus individuals were exposed in the field might have added a large variation to our morphological results. Additionally, within some populations, variation between individuals was high, potentially swamping differences across populations. Moreover the widespread occurrence of hybrid populations makes identification using morphological data even more difficult (Suehs et al. 2004).

Despite the challenges related to the morphological identification of invasive Carpobrotus species using morphological characters, our results have important implications for the development of management programmes. First, no introductions of individuals from cluster C have been detected in any of the sampled sites. However, the rate of introduction of alien species is rapidly increasing (Seebens et al. 2021). This, coupled with the widespread use of Carpobrotus species as ornamental plants, enhances the chances of individuals from cluster C to be introduced and the potential for genetic exchange between populations from all three clusters, which could increase the invasion success of Carpobrotus. Hence, management strategies should aim at preventing the introduction of additional Carpobrotus genotypes, especially from the Eastern South African coast. Second, we revealed that the most probable sources of Carpobrotus introductions and invasions globally are the Western Cape province in South Africa and California. Since most effective biocontrol agents are generally those that have co-evolved with the invasive species (Müller-Schärer et al. 2004), the search for biocontrol agents to manage Carpobrotus invasions should be focused in these areas. A challenge for this in California is the fact that pure C. chilensis is rare due to the extensive hybridization, and no specialist insects have been observed on it other than the rare occurrence of the introduced scale insects Pulvinariella mesembryanthemi and Pulvinaria delottoi (Schmalzer and Hinkel 1987). Moreover, there is no clear evidence that C. chilensis is native to California, and future studies should extend sampling efforts to other areas such as the coasts of Chile and Australia. Additionally, it should be carefully explored whether biocontrol agents from the Western Cape province in South Africa and California are effective at managing admixture (or hybrid) populations or Carpobrotus invasions in general.

Our work highlights exciting opportunities for future research on Carpobrotus invasions. For example, high-resolution population genomic analyses (e.g., single nucleotide polymorphism genotyping or whole genome sequencing), coupled with common garden experiments, would provide valuable insights into the diversity and evolutionary dynamics of the genus, the invasiveness of its representatives and their interactions with insects with the potential to be used for biological control. For instance, a highly flexible breeding system that allows extensive hybridization (i.e., outcrossing) and high levels of clonal reproduction (via vegetative structures) suggest the stabilization of highly successful hybrid genotypes is likely to occur. Determining whether certain hybrid combinations and/or clones are more prevalent in native or invasive ranges should be included in future research to inform future management of the group.

AN, MLC, SC, JR and JJLR acknowledge funding by project no. 19-13142S, and PP by EXPRO grant no. 19-28807X (both from Czech Science Foundation). SC was supported by the Irish Research Council COALESCE/2021/117 award. JR acknowledges funding from the Spanish Ministry of Universities under application 33.50.460A.752 and the European Union NextGenerationEU/PRTR through a contract Margarita Salas of the Universidade de Vigo (UP2021-046). AN, MLC, SC, JR, DMR and PP were supported by a long-term research development project no. RVO 67985939 (Czech Academy of Sciences). DMR also received support from Mobility 2020 project no. CZ.02.2.69/0.0/0.0/ 18_053/0017850 (Ministry of Education, Youth and Sports of the Czech Republic). CMD thanks L. Lum for access to federal govt land for plant collections in California.