Giglio Island (WGS84: 42.35527°N, 10.90134°E) is the second largest island of the Tuscan Archipelago (Italy) with a surface area of 21.2 km² and a perimeter of 28 km (Baldini 1998). Almost all Giglio Island’s land surface, approximately 21 km², falls within the Natura 2000 Special Area of Conservation (SAC/SPA IT51A0023), with a portion of it, covering 8.9 km², being part of the Tuscan Archipelago National Park.

The island is mainly mountainous, with steep and rocky slopes up to the coastline. The climate of Giglio Island is Mediterranean, with mild, rainy winters and hot-arid summers, peaking in July and August (Baldini 1998; Foggi and Pancioli 2008). The vegetation is typically Mediterranean, with forests dominated byQuercus ilex, evolved scrubs of Erica arborea and Arbutus unedo, and other typical Mediterranean Cistus and Helichrysum garrigues (Foggi and Pancioli 2008). The coastal vegetation, of particular interest for this study, being the one invaded by Carpobrotus (see Fig. 1A, B), hosts a mosaic of habitats of conservation interest according to Directive 92/43/EEC “Habitat”, including the habitat of Vegetated sea cliffs of the Mediterranean coasts with endemic Limonium spp. (habitat code 1240 according to Dir. 92/43/EEC, hereafter referred to as Vegetated sea cliffs), and the habitat of Low formations of Euphorbia close to cliffs (hab. code 5320, hereafter referred to as Coastal garrigues). Vegetated sea cliffs are characterized by the presence of alophylous chasmophytes such as the endemic Limonium sommierianum Fiori and Crithmum maritimum L, with few other species very sporadic (f.e. Lotus cytisoides L., Catapodium pauciflorum (Merino) Brullo, Giusso, Miniss. & Spamp. or Polypogon subspathaceus Req.). This paucispecific habitat hosts species fully adapted to grow on rocky coasts and occupy the first colonizable zone in contact with the sea, forming a discontinuous belt along the entire perimeter of the island. Indeed, Vegetated sea cliffs are generally characterised by an almost total absence of soil and direct contact with seawater and marine aerosol, therefore requiring highly specialised species (Perrino et al. 2013). Coastal garrigues settle in the upper parts of rocky coasts, in a variable-width zone between the typical aeroalophilous vegetation of the Vegetated sea cliffs and the first elements of Mediterranean low and high maquis. These plant communities are characterized by a low, single-layer structure dominated by dwarf and small shrubs like Helichrysum litoreum Guss. and Jacobaea maritima (L.) Pelser & Meijden., prostrate Pistacia lentiscus L., and host several other Mediterranean annual species.

Figure 1.

Some images of the interventions of Carpobrotus removal at Giglio Island A, B rocky cliffs invaded by Carpobrotus. In 2020 C, D images of the cliffs after the main intervention in early 2022 E, F monitoring plots after one year (2022) with the visible presence of Carpobrotus seedlings and G in 2023 with native vegetation.

The present study analysed the changes in Carpobrotus and native species presence and abundance during the actions of control of this invasive alien species, on the island of Giglio conducted within the EU LIFE project LIFE LETSGO GIGLIO “Less alien species in the Tuscan Archipelago: new actions to protect Giglio island habitats” (LIFE18 NAT/IT/000828). In the spring and summer of 2020, we conducted a preliminary phase of detailed mapping of the spread of Carpobrotus on the island. This involved the interpretation of aerial photos and surveys on the island. The initial distribution of this species on the island was recorded as approximately 61,000 m², with the majority occurring on cliffs or rocky areas. Approximately 50,000 m² of this area was found to be strictly invaded (Lazzaro et al. 2016; Mugnai et al. 2022). The intervention area extended over 33000 m² of coastal habitats, with an estimated net Carpobrotus surface of approximately 22000 m².

The main intervention for the removal began in the winter of 2021–2022 integrating two techniques: manual removal and covering with mulch sheets (landscape fabric, 105 g/m²). It should be noted that approximately 90% of the Carpobrotus on the island were removed manually, while mulching sheets were used only in a limited number of areas, mostly used to contain and isolate the waste material (Fig. 1C, D). The manual removal involved the entire plant, including roots, with the objective of eliminating all visible Carpobrotus live plants within the intervention areas. As expected, regrowth of seedlings occurred in all the areas subjected to treatment, particularly in areas where the seed bank was present in the litter remaining on the ground (Fig. 1E, F). Hence, the project foresees continuous monitoring for three years following the main intervention and several rounds of follow-up interventions for the removal of new seedlings. Indeed, already in the first two years after the main intervention (from 2022 to 2023), any seedlings and resprouts were removed manually in April/May and, at the time of writing, continue to be removed annually (Fig. 1G). Furthermore, monitoring and the removal of seedlings will continue for at least five years after the main intervention.

Further technical details on the methods adopted are available in the executive project for the eradication (https://www.lifegogiglio.eu/wp-content/uploads/WEB_Relazione-illustrativa-generale_Carpobrotus-1.pdf).

The sampling was conducted at the promontory of “Punta Capel Rosso”, south of the island, largely invaded by Carpobrotus. In particular, the monitoring was carried out on both habitats, Vegetated sea cliffs and Coastal garrigues, as they were the two most invaded habitats of major conservation importance within the study area.

The monitoring began in 2020, and the impact assessment was carried out using the Before-After-Control-Impact (BACI), a suitable evaluation scheme consisting of pre- and post-intervention sampling of the restoration sites and control sites (Christie et al. 2019). The experimental sampling design has been implemented and maintained through a floristic survey of 2 × 2 m squared plots in two treatments: invaded (plots subjected to the removal) and control (plots in native vegetation). The sampling was stratified according to a random sampling design based on the surface of the habitats mapped concordant to the HaSCITu (Habitat in the Sites of Conservation Interest in Tuscany) program, and a detailed mapping of the distribution of Carpobrotus confirmed after several visits to the island before the commencement of the monitoring. Plots were paired, thus for each invaded plot a control plot was identified as close as possible. The sampling included a total of 24 permanent plots, 6 × 2 plots for Vegetated sea cliffs, and 6 × 2 plots for Coastal garrigues, evenly distributed between invaded and control.

Vegetation sampling was carried out during the vegetative season in May when most species were identifiable. Each plot was georeferenced and marked with a peg and a numbered nameplate. Data on native plant species occurrence and abundance was collected using a percentage scale, considering the overlapping of different species. Furthermore, the percentage of fresh Carpobrotus as well as its dead litter, was recorded. We present the data collected up to 2023, including thus 4 years of monitoring, resulting in the survey of 96 plots. A full list of species observed during the sampling within the two habitats between 2020 and 2023, is included in Suppl. material 1.

To verify the effect of Carpobrotus removal intervention on its cover and litter, and on the alpha diversity of native vegetation, we used a Repeated Measurement ANOVA-type modelling. We fitted a series of linear mixed models, with a random effect factor on plot identity to account for the autocorrelation linked to the repetition of the samplings across the four years of surveying and including a covariance structure based on a Gaussian spatial autocorrelation of the observations accounting for the paired structure of the sampling design. For each of the two habitats separately, we assessed whether the cover and litter of Carpobrotus varied according to treatment and time. Similarly, we tested whether native species richness (SR), native diversity expressed as H’ index, and native species abundance, expressed as the sum percentage cover of each species, changed before, during, and after the interventions using time and treatment (Invaded vs. Control) as fixed effect explanatory variables. When required, the variables were log or asin-transformed to achieve the normality of residuals.

We assessed the changes in the species composition of plots using multivariate analysis for the two habitat types. The analyses included only plots with at least one species (a total of 92 plots; 4 plots had no species in 2022, the year of main interventions). Plot species composition differences were analysed using a non-metric multidimensional scaling (NMDS) analysis based on Bray-Curtis dissimilarities calculated on abundance data (expressed as percentages). Nitrophilous species were defined as those with Ellemberg’s ecological indicator value for eutrophication above 6 (Pignatti 2005; Ellemberg 2009), the relative abundance and frequency of nitrophilous species were calculated and, due to the differences in the two types of plant communities, we conducted all the analyses separately for the two habitats. Moreover, we evaluated the extent and the divergence of species composition variations according to time and treatment using a Principal Response Curve (PRC) analysis (ter Braak and Smilauer 2012). In this analysis, time was treated as a categorical variable and was used as a covariate. The significance of the effect of the treatment on the species composition was assessed with a permutation analysis using 9999 permutations due to the hierarchical structure of the data and allowing freely exchangeable permutation on the whole plot level, and no permutation at the split-plot level. Given the differences in the two types of plant communities within the two habitats, we conducted the PRC analysis separately for the two habitats.

All analyses were conducted in the R environment (R version 2023.06.2): the LME models were fitted using the ‘nmle’ package version 3.1-162 (Pinhero and Bates 2006); the NMDS was produced using the ‘vegan’ package version 2.6-4 (Oksanen et al. 2020); PRC analysis was performed using ‘prc’ function in the ‘vegan’ package version 2.6-4, (Oksanen et al. 2020). All plots were drawn using ‘ggplot2’ package version 3.4.2. (Wickham 2016).

Our results build on the effects of Carpobrotus removal on coastal reef plant communities, based on a four-year survey period, focusing on describing and analysing the short-term response of native vegetation. Our results markedly indicated that the changes in community composition through the years correspond to a prompt recovery of the native plant communities following the removal of Carpobrotus. Prior to removal, both Vegetated sea cliffs and Coastal garrigues invaded communities were similar in composition, due to the very high impact of Carpobrotus at the alpha diversity level replacing and outcompeting characteristic native species as already shown in several similar contexts (Fried et al. 2014; Badalamenti et al. 2016; Mugnai et al. 2022). However, invaded plots of both types tended to converge towards their respective control plots during the following growing season after the removal of Carpobrotus. Following the removal, the Vegetated sea cliffs communities, were primarily colonised by Jacobaea maritima subsp. maritima, Lotus cytisoides and Limonium sommierianum, whereas those of Coastal garrigues were characterised by Helichrysum litoreum, and Pistacia lentiscus, which are indicator species of the typical coastal vegetation (Foggi and Pancioli 2008). Moreover, some of these species have a high conservation and naturalistic value, such as Limonium sommierianum, as it is endemic to the Tuscan Archipelago (Baldini 1998). Similarly, on Bagaud Island (Var, France), following the removal of Carpobrotus, the post-eradication plant communities of rocky coastal cliffs exposed to sea spray tended to converge toward baseline communities characterised by species such as Jacobaea maritima, Spergularia spp. and Polycarpon tetraphyllum (Krebs et al. 2015). Hence, our data confirm once more the effectiveness of control actions in the case of Carpobrotus invasion. Indeed, the effectiveness of manual removal and mulching sheet on the removal of Carpobrotus has been already shown (see Lazzaro et al. 2023; Nascimento et al. 2023), but while other experiences have shown that the recovery of native communities can be achieved on a medium/long term (7 and 5 years after the interventions, respectively in Buisson et al. 2021 and Lazzaro et al. 2023), we show that already in two years native plants can colonize the areas where Carpobrotus had been manually removed.

Although the removal of Carpobrous leads to an inevitable decrease in Carpobrotus coverage and a slow increase in the recolonization of native plants, the persistence of litter in situ can lead to a higher potential for reinvasion due to its large seed bank (Chenot et al. 2014). Indeed, as expected, in two growing seasons after removal, the emergence of seedlings of Carpobrotus has been observed within the survey plots and in all the study areas and appeared greater in the coastal garrigues and where litter was thicker and persistent after the interventions. Novoa et al. (2012, 2013) have shown that litter accumulation can repress the growth of native plants even after the removal of live plants, limiting the suitable area for the recolonisation of the natives. However, removing the litter, other than adding a significant cost to the removal operation, may leave the soil subject to erosion (Chenot et al. 2018). On coastal reefs as per their exposure to atmospheric elements, the persistence of litter decreases quickly: we noted that litter cover diminished in 2023, probably due to soil runoff caused by wind and heavy autumn rains. Interestingly, with an opposite trend, litter was found in some control plots in Vegetated sea cliffs, with no live Carpobrotus present, probably transported from the invaded plots to the control plots during bad weather.

However, in one case, it has been shown that leaving the litter improves the germination of new Carpobrotus seedlings (Chenot et al. 2018) while, in another, the removal of both the living parts and the litter guarantees a more effective restoration of the native vegetation (Novoa et al. 2013), our results showed a good recovery of the native vegetation a few years after removal with a trend that increasingly approaches that of the control plots. In line with our results, Buisson et al. (2021) pointed out that coastal vegetation plant communities can recover and become relatively similar to the reference within a few years, while shrub communities may need more time to recover due to the competition from native herbaceous species. Buisson et al. (2021) and Campoy et al. (2018) suggested that Carpobrotus may have a persistent seed bank (> 5 years), but there are no exhaustive studies over a longer period. Gioria et al. (2012) placed Carpobrotus edulis in the category of short-term persistent seed banks, and Ruffino et al. (2015) reported that seeds of Carpobrotus can persist for 5 years after eradication. Numerous studies (Novoa et al. 2013; Ruffino et al. 2015; Buisson et al. 2021; Lazzaro et al. 2023) emphasised that it is, therefore, necessary to carry out regular monitoring of treated areas for 10 years before being able to assess the success of the eradication.

There are strong differences between Vegetated sea cliffs and Coastal garrigues habitats in terms of species richness and diversity (Foggi and Pancioli 2008; Mugnai et al. 2022), native species cover, and the response of their native communities to the removal of Carpobrotus. Moreover, the rapid response of the native vegetation follows the trend described by Buisson et al. (2021), Lazzaro et al. (2020a), and Andreu et al. (2010), confirming that, in the absence of post-removal disturbance, invaded and then treated areas can quickly and naturally revegetate, reaching levels of species richness and species diversity close to those of the corresponding control areas. Indeed, in the year following the removal of Carpobrotus, as shown in Fig. 2, species richness and diversity values in treated plots for Vegetated sea cliffs exceeded those of the control. This is likely due to the accumulation of Carpobrotus litter in areas that normally would not experience such soil buildup, which has allowed for the establishment of richer communities than those typically found in Vegetated sea cliffs. In 2023 there is also a slight divergence at the beta-diversity level, probably due to the entry of some ruderal and nitrophilous species, that are not typically found in either of these habitats, but which can be favoured by soil nutrient enrichment caused by Carpobrotus (Novoa et al. 2013; Fried et al. 2014; Malavasi et al. 2016; Lazzaro et al. 2023). The nitrophilous species with the highest cover values in both 2022 and 2023 are Sonchus asper (L.) Hill, Polycarpon tetraphyllum (L.) L., Polypogon subspathaceus, Mercurialis annua L. and Dactylis glomerata L. all with Ellemberg’s ecological indicator values fluctuating between 6 and 8. Similarly, also for Bagaud Island (Var, France), following the removal of Carpobrotus, there was a shift of the post-eradication plant communities of the inner part of the island to the reference alonitrophilous ones. Although this inland area already had alonitrophilous herbaceous communities due to the presence of Larus michahellis (Naumann, 1840), there was no post-eradication community shift to low matorral communities, which are also present in this area (Krebs et al. 2015). Indeed, both Carpobrotus and Larus michahellis have been demonstrated to facilitate soil enrichment and entry of nitrophilous species (Novoa et al. 2013; Krebs et al. 2015).

Given the specialised flora that usually characterize the Vegetated sea cliffs, the occurrence of nitrophilous species in this habitat could probably be due to an unusual accumulation of Carpobrotus litter, which may have allowed the expansion of some non-characteristic species. In contrast, the relative abundance of nitrophilous species is reduced in habitat 5320 due to the greater complexity of this habitat, which develops on soils between the cliffs exposed to the action of the sea and the shrub communities of the more internal thermo-Mediterranean scrub (Perrino et al. 2013).

The authors have declared that no competing interests exist.

No ethical statement was reported.

The research was funded by the LIFE Nature and Biodiversity project LIFE LETSGO GIGLIO “Less alien species in the Tuscan Archipelago: new actions to protect Giglio island habitats” (LIFE18 NAT/IT/000828), and by the Italian Ministry of University and Research, PNRR, Missione 4 Componente 2, “Dalla ricerca all’impresa”, Investimento 1.4, Project CN00000033.

Alice Misuri: Methodology; Investigation; Visualization; Writing – Original Draft. Eugenia Siccardi: Methodology; Investigation; Visualization; Writing – Original Draft; Formal Analysis; Data curation. Michele Mugnai: Conceptualization; Methodology; Investigation; Visualization; Writing – review & editing. Renato Benesperi: Project administration; Validation; Writing – review & editing. Francesca Giannini: Project administration; Validation; Writing – review & editing. Michele Giunti: Project administration; Methodology; Validation; Writing – review & editing. Lorenzo Lazzaro: Project administration; Conceptualization; Methodology; Investigation; Writing – review & editing; Funding acquisition; Validation.

Alice Misuri https://orcid.org/0009-0006-5352-0845

Eugenia Siccardi https://orcid.org/0009-0008-4738-0633

Michele Mugnai https://orcid.org/0000-0003-4315-2920

Renato Benesperi https://orcid.org/0000-0003-4296-3393

Lorenzo Lazzaro https://orcid.org/0000-0003-0514-0793

All of the data that support the findings of this study are available in the main text or Supplementary Information.