The study area encompasses a representative coastal landscape of Mediterranean Holocene dunes on the Tyrrhenian coast in Central Italy (Lazio Region) of approximately 280 km (Fig. 1a; Mazzini et al. 1999; Amato et al. 2012). These dunes that are low and narrow and occupy a 400–500 m wide strip parallel to the shoreline (Acosta et al. 2003), in natural conditions, host a well-developed vegetation zonation, which follows a steep sea-inland abiotic gradient from pioneer communities dominated by annual plants to the inner sectors of Mediterranean scrub (Acosta et al. 2003; Fig. 1b). In addition to alien plant invasions (Bazzichetto et al. 2018a), the analysed dunes are highly threatened by coastal erosion (Bazzichetto et al. 2020), urban expansion (e.g. new buildings, recreational sites etc.), land take (e.g. agricultural expansion, afforestation, industrial and harbour development etc.) and ecosystems fragmentation (Malavasi et al. 2014, 2018a).

Figure 1.

a Map of the study area (reference system WGS84 UTM 33 N, EPSG: 32633) showing the distribution of the invaded dune systems b schematic profile of coastal dune zonation with the acronym of the mapped natural coastal classes: SEA – sea water, BPV – beach with pioneer vegetation, HDV – herbaceous dune vegetation, SWV – Shrub woody vegetation, FWV – forest woody vegetation.

Carpobrotus acinaciformis (L.) L.Bolus and C. edulis (L.) N.E.Br. are mat-forming trailing succulent perennial herbs that are native to South Africa (Wisura and Glen 1993). Since both species tend to invade similar coastal dune habitats with comparable behaviour and impacts (Sarmati et al. 2019) and they commonly hybridise with each other, in their invasive range, they are frequently referred to as Carpobrotus spp. (Novoa et al. 2023). In the study area, Carpobrotus acinaciformis (L.) L.Bolus, C. edulis (L.) N.E.Br. and probably also hybrids between the two, were introduced as ornamental plants and for consolidating dunes (Campoy et al. 2018). As on other sectors of their invasive range, Carpobrotus spp. tend to invade herbaceous dune vegetation growing on shifting and fixed dunes preferentially and, to a lesser extent, clearings in shrubland and understoreys in woody vegetation and fore dunes (Carranza et al. 2011; Bazzichetto et al. 2018b).

The workflow for analysing changes in invaded coastal landscapes and for assessing the interplay between invaded patch dynamics and contextual landscapes is illustrated in Fig. 2. It includes four key steps: (A) bi-temporal Carpobrotus spp. and land-cover mapping, (B) coastal dune landscape change, (C) spatial pattern analysis over time and (D) relationship between IAP dynamics and coastal landscape pattern.

Figure 2.

Schematic overview of the procedure implemented to analyse the spatiotemporal changes of Carpobrotus spp. invasion in relation with the landscape context. A Bi-temporal mapping of Carpobrotus spp. and land cover B analysis of landscape change C temporal spatial pattern analysis D assessment of the relationship between IAP dynamics and coastal landscape pattern.

Spatial pattern analysis over time

Changes in the spatial pattern of the coastal tracts invaded by Carpobrotus spp. were assessed by calculating and comparing, over time, a comprehensive set of landscape metrics (LM) that depict spatial composition and configuration at both the landscape level (LM_land) and the class level (LM_class) for the different time steps (T0 and T1; Fig. 2C; Table 2; Riitters et al. (1995)). We selected metrics able to depict distinct ecological processes within dune landscape patterns, whether these processes impede or facilitate plant invasion. These metrics may exhibit non-linear correlation issues and encompass key ecological processes and mechanisms essential for advancing the ecological understanding of Carpobrotus spp. pattern of change (Smith et al. 2009; Long et al. 2010). In particular, the selected metrics capture critical drivers of the invasion process on coastal dunes, including dune fragmentation and urban expansion (Malavasi et al. 2014, 2018a; Carranza et al. 2015), erosion/accretion dynamics (Bazzichetto et al. 2020), the loss of integrity within natural dune mosaics (Acosta et al. 2003) and the presence of both natural and artificial corridors facilitating alien propagule dispersal (Bazzichetto et al. 2018b).

Table 2.

Download as

CSV

XLSX

Names (acronyms), formulas, descriptions, units of measurement and the associated spatial pattern levels (Class/Landscape) and facets (composition/configuration) of the selected pattern metrics. A = total landscape area, nj = number of patches of j-land-cover class, aij = area of the i-th patch of j-land-cover class, eij = total length of the i-th patch edge of j-land-cover class, m = total number of land-cover classes, Pj = proportion of the landscape occupied by j-land-cover class.

Name (Acronym)	Formula	Description	Unit / Range	Pattern facet
Class level (LMclass)
Percentage of Landscape (PLAND)	$\frac{{\sum }_{i=1}^{n_j}{a}_{ij}}{A}*100$	Sum of the areas (m2) of all patches of the j-land cover class, divided by coastal tract area (m2) in percentage. Measure of dominance/rareness.	Percent (%) 0 ≤ PLAND < 100	Composition
Patch density (PD)	$\frac{n_j}{A}$	Density of patches of the j-land-cover class per unit area. Measure of aggregation/dispersion.	Number/ha PD > 0, no limit.	Configuration
Edge density (ED)	$\frac{{\sum }_{i=1}^{n_j}{e}_{ij}}{A}$	Edge length of j-land-cover class on the landscape area. Length of the contact with other classes. Measure of shape complexity/simplicity.	Metress/ha ED > 0, no limit	Configuration
Mean patch area (AREA_MN)	$\frac{{\sum }_{i=1}^{n_j}{a}_{ij}}{n_j}$	Area of j-land-cover class divided by its number of patches. Measure of fragmentation/colonisation.	Ha AREA_MN > 0, no limit	Configuration
Landscape level (LMland)
Shannon Diversity Index (SHDI)	$-\sum _{i=1}^m\left(P_i*\ln P_i\right)$	Shannon’s Diversity Index accounting for land-cover class richness and equitability. Sensitive to rare land-cover classes.	Natural number 0 ≤ SHDI < ∞ SHDI = 0 – no diversity	Composition
Simpson Diversity Index (SIDI)	$1-\sum _{j=1}^m{P}_j^2$	Simpson’s Diversity Index depicting land-cover class richness and dominance. Sensitive to dominant land-cover classes.	0 ≤ SIDI <1. SIDI = 0 – no diversity	Composition

For landscape metrics (LM_land), we calculated and compared over time two indices depicting landscape richness (number of land-cover classes) and evenness (relative abundance of each class). As Shannon index (SHDI), including a logarithmic transformation of abundance values (Table 2), is particularly sensitive to less abundant classes and constitutes a good indicator of equipartition, Simpson’s index (SIDI; Table 2), is a reliable measure of dominance (McGarigal et al. 2012).

For class metrics (LM_class), we calculated and compared over time four indices, illustrating spatial composition (class abundance) and configuration (class spatial pattern, see Table 2; McGarigal et al. (2012)). Amongst these class metrics, PLAND, representing the percentage of the landscape covered by a given class, is a good surrogate for class dominance or rareness. PD, or patch density, describes the number of patches per unit area and measures class aggregation or dispersion in the landscape. ED, or edge density, calculated as the ratio between class perimeter and landscape area, measures class shape complexity or simplicity (e.g. edge effects vs. core areas) and provides an overview of the contact of a class with other land cover classes (e.g. high values may suggest the role as landscape matrix). AREA_MN, measured as the mean patch dimension of a given class, depicts the presence of large or small patches and is a good indicator of natural habitats fragmentation (e.g. smaller patches of native vegetation), as well as of invasive plants colonisation (e.g. increasing size of alien species patches over time; Table 2; McGarigal et al. (2012)). Metrics were calculated with FRAGSTAT 4.2 software (McGarigal et al. 2012).

Spatial pattern changes over time at class level (LM_classT0 vs. LM_classT1) were analysed by trajectory analysis (sensu Long et al. (2010)) depicting the relationship between composition and configuration metrics (Carranza et al. 2015; Malavasi et al. 2018b). For each class, a specific bi-temporal relationship space was produced by projecting in a Cartesian diagram the class configuration metric values (PD, ED, AREA_MN) computed for each coastal tract against the respective percentage of class cover (PLAND). Then, the arithmetic mean of each class metric (mean_LM_class) was plotted in the relationship space and the temporal trajectories for EXP_CAR and RED_CAR were drawn by connecting metric means chronologically with arrows. After a visual inspection of the relationship space to ascertain the means can be statistically compared, we assessed pattern landscape and class metric changes (LM_classT0 vs. LM_classT1 and LM_landT0 vs. LM_landT1) by the non-parametric pairwise Wilcoxon rank test.

The temporal changes in composition given by the mean values of PLAND (e.g. in PLAND_CAR, PLAND_HDV, PLAND_ART etc.) and configuration metrics assessed as the mean of PD, ED and AREA_MN (e.g. in PD_CAR, ED_HDV, AREA_MN_ART etc.) were interpreted accounting of the specific non-linear relationship amongst them (Fig. 3; Long et al. (2010)). Given that numerous previous studies have demonstrated the non-linear correlation between landscape composition (PLAND) and configuration metrics (PD, ED, AREA_MN) across different environments, such as grasslands, forests, croplands and urban areas (Long et al. 2010; Su et al. 2012; Carranza et al. 2015; Malavasi et al. 2018b; Hermosilla et al. 2019; Zhang et al. 2020), it is essential to simultaneously analyse their temporal changes (Fig. 3). Such a concurrent interpretation of pattern metrics facilitates a nuanced understanding of the intricate dynamics interweaving these landscape facets (Long et al. 2010; Carranza et al. 2015).

Figure 3.

Combined effects of temporal changes (from T0 to T1) in composition (e.g. increasing or decreasing PLAND values) and configuration (e.g. increasing or decreasing PD, ED, AREA_MN values) of a hypothetical land-cover class: a changes in PLAND and PD b changes in PLAND and ED c changes in PLAND and AREA_MN.

In general, as described by Long et al. (2010), an increase in patch density (PD) accompanied by a decrease in PLAND values (e.g. numerous smaller patches) may indicate the fragmentation of natural habitats (Fig. 3a, Carranza et al. (2015)). Conversely, a simultaneous rise of PD and PLAND (e.g. numerous, larger patches) may reflect colonisation processes (Fig. 3a). Furthermore, a decrease in PD could signify either increasing fragmentation (if accompanied by a reduction in PLAND) or habitat expansion due to the aggregation of several small patches into fewer, larger ones (if accompanied by PLAND increase; Yang and Mountrakis (2017)). Similarly, a rise in ED values can correspond to two scenarios (Fig. 3b). A simultaneous increase of ED and PLAND (e.g. larger and more irregularly-shaped patches) is likely associated with habitat frontal expansion. On the other hand, an increase in ED combined with a decline in PLAND (e.g. smaller irregularly-shaped patches) may indicate habitat fragmentation caused by patch shrinkage and irregular edge erosion (Carranza et al. 2015). On the other hand, decreased ED values can be associated with fragmentation if accompanied by a reduction in PLAND or with habitat expansion and the aggregation of irregular patches into larger, more regular ones if accompanied by an increase in PLAND (Yang and Mountrakis 2017). With regard to AREA_MN (Fig. 3c), a decline in average patch size combined with increasing PLAND may indicate habitat expansion with the formation of new small patches. Conversely, the simultaneous reduction of AREA_MN and PLAND (e.g. smaller patches) suggests fragmentation of larger patches into smaller ones, along with a reduction in the cover of the remaining patches (Long et al. 2010; Carranza et al. 2015). An increase in AREA_MN and PLAND (e.g. lager patches and higher dominance in the landscape) may indicate habitat expansion. In contrast, an increase in AREA_MN accompanied by a reduction in PLAND reduction (e.g. large residual patches) may depict habitat loss, characterised by the disappearance of several medium-to-small patches and the persistence of one or few large patches (Malavasi et al. 2018b).

The produced land-cover maps are highly accurate (OA, K, TSS and BA greater than 85.682%, 0.833, 0.773 and 0.886, respectively; Suppl. material 1: tables S2, S3) and, thus, reliable for further landscape analysis. A comparison of land-cover maps between T0 and T1, using transition analysis, revealed changes in the 18% of the landscape, with most land-cover classes shifting towards the neighbouring categories in a comparable manner (Fig. 4a–c; Suppl. material 1: table S4).

Figure 4.

Chord diagrams for: a all coastal tracts b tracts with Carpobrotus spp. expansion (EXP_CAR) and c tracts with Carpobrotus spp. reduction (RED_CAR). The chord diagrams summarise the percentage (%) of each land-cover class in T0 (outer ring) that changed into another class to T1. The size and the direction of arrows represent transitions to other classes in T1. For example, an increase of HDV from BPV in T1 considering all coastal tracts. The proportion (%) of each land-cover class that remained stable over time is represented by the internal coloured circle. Land-cover classes: artificial areas (ART), beach with pioneer annual vegetation (BPV), herbaceous dune vegetation (HDV), shrub dune vegetation (SDV), Forest and woody dune vegetation (FDV), semi-natural herbaceous vegetation (SHV), wetland (WET), Tyrrhenian sea (SEA), Carpobrotus spp. (CAR).

In both years, the dominant categories are Artificial areas (ART) and Open Sand (BPV) summing up to over the 40% of the mapped area, followed by herbaceous dune vegetation (HDV) and Sea (SEA) covering over 12% (Fig. 4a; Suppl. material 1: table S4). In EXP_CAR coastal tracts, the landscape was relatively stable with weak accretion processes (reduction of SEA class area; Fig. 4b), while in RED_CAR, landscape tracts have changed with a consistent increase in the SEA category at the expense of BPV (Fig. 4c; Suppl. material 1: tables S5, S6). Four cover classes (SDV, SHV, WER, FDV), which had limited extension in both time steps, were excluded from further analysis and modelling.

The chord diagram of the overall landscape evidences balanced shifts between CAR and HDV (Fig. 4a) differently, the separate analysis of chord diagrams revealing opposite landscape dynamics on EXP_CAR and RED_CAR. In EXP_CAR areas, we observed the expansion of CAR class at the expense of the HDV class (Fig. 4b), while in RED_CAR landscapes, we registered HDV replacing Carpobrotus spp. class; Fig. 4c).

The spatial pattern of Carpobrotus spp. patches changed significantly over time (from T0 to T1) presenting opposite trends in EXP_CAR and RED_CAR tracts (Fig. 5). In EXP_CAR tracts, Carpobrotus spp. patches increased in extent ranging from 0.41 m² to 865.84 m². Conversely in RED_CAR tracts, the extent of Carpobrotus spp. decreased ranging from – 1.65 m² to – 1322.25 m² (Suppl. material 1: fig. S10).

Figure 5.

Comparison of Carpobrotus spp. pattern metrics (PLAND_CAR, PD_CAR, ED_CAR, AREA_MN_CAR) over time (from T0 to T1) in tracts of expansion (red: EXP_CAR) and reduction (blue: RED_CAR). a Kruskal-Wallis comparison of means and the respective confidence intervals (upper – U. CI, lower – L. CI; * = p-value < 0.05, ** = p-value < 0.01, *** = p-value < 0.001) b–d report the trajectory analysis of Carpobrotus spp. patches area (AREA_MN_CAR), edge density (ED_CAR) and patch density (PD_CAR) in relation to overall Carpobrotus spp. cover (PLAND_CAR). Grey dots represent the observed values of pattern metrics, coloured dots the arithmetic mean (mean of PLAND_CAR, PD_CAR, ED_CAR, AREA_MN_CAR) in each date (T0 and T1) in EXP_CAR and RED_CAR and arrows indicate the direction of temporal change.

In tracts of IAP expansion (EXP_CAR), all Carpobrotus spp. spatial metrics significantly increased. The extension of invaded areas increased (greater PLAND_CAR, Fig. 5a–d) with Carpobrotus spp patches distributed on larger (AREA_MN_CAR, Fig. 5a, b), more numerous (PD_CAR, Fig. 5a, d) and irregularly-shaped patches (ED_CAR, Fig. 5a, d). On EXP_CAR landscape, we also registered a significant increase of artificial patch dimension (AREA_MN_ART; Suppl. material 1: fig. S11). On the other hand, in coastal tracts of IAP reduction (RED_CAR), the decline of Carpobrotus spp. (PLAND_CAR; Fig. 5a–d) coincides with simpler configuration metrics, indicated by lower values in AREA_MN_CAR (Fig. 5a, b) and ED_CAR (Fig. 5a, c).

Changes on Carpobrotus spp. pattern (∆LM_CAR) are related with landscape changes (∆LM_land), specifically concerning sea (SEA), herbaceous dune vegetation (HDV), the artificial surfaces (ART) classes (Figs 6, 7, Suppl. material 1: table S7). The Partial Dependence plots (Figs 6, 7) evidence that such relationship varies amongst areas of Carpobrotus spp. expansion (EXP_CAR) and reduction (RED_CAR). The RF models and their setup (Mtry, Ntree, MNS) provided an adequate description of the landscape dynamics of Carpobrotus spp. invasion, including changes in the spatial pattern of the surrounding landscape (∆LM_class and ∆LM_land; Suppl. material 1: table S8). The spatial-temporal dynamics of Carpobrotus spp. patches (∆PLAND_CAR, ∆PD_CAR, ∆ED_CAR, ∆AREA_MN_CAR) are significantly correlated with coastal dune landscape changes (with a minimum R² of 0.405 for the EXP_∆PD_CAR model and 0.595 for the RED_∆PLAND_CAR model; Figs 6, 7).

Figure 6.

Partial dependence plots (PDP) using linear smoothing of the most important variables (over of 40% in cumulate importance) on the RF models for areas of alien expansion (EXP_∆LM_CAR). a EXP_∆PLAND_CAR b EXP_∆PD_CAR c EXP_∆ED_CAR d EXP_∆AREA_MN_CAR. Red dotted lines represent raw PDP curves. The importance of each variable is indicated by the Mean Decrease Importance (MDI) value. For land-cover classes, see Table 1 and for pattern metrics description, see Table 2.

Figure 7.

Partial dependence plots (PDP) using linear smoothing of the most important variables (over of 40% in cumulate importance) on the RF models for areas of alien reduction (RED_∆LM_CAR). a RED_∆PLAND_CAR b RED_∆PD_CAR c RED_∆ED_CAR d RED_∆AREA_MN_CAR). Red dotted lines represent raw PDP curves. The importance of each variable is indicated by the Mean Decrease Importance (MDI) value. For land-cover classes, see Table 1 and for pattern metrics description, see Table 2.

In coastal tracts of IAP expansion (EXP_CAR), the landscape change variables (RF_∆LM) that better explain Carpobrotus spp. evolution (EXP_∆LM_CAR) are the size and the shape complexity of herbaceous dune vegetation, of the sea and of artificial areas (∆AREA_MN_HDV, ∆AREA_MN_SEA, ∆AREA_MN_ART, ∆ED_HDV, ∆ED_SEA, ∆ED_ART; Fig. 6). A higher increment of Carpobrotus spp class cover (∆PLAND_CAR) occurs in correspondence with the reduction of sea-class surface (∆AREA_MEAN_SEA ≈ -0.25 to -0.50), with increasing contacts with artificial infrastructures (∆ED_ART ≈ 100 to 250) and with intermediate reduction in the number of natural dune vegetation patches (∆PD_HDV ≈ -150). We also registered an increase in the number of invaded patches (∆PD_CAR, Fig. 6b) in correspondence with increasing edge length of beach-pioneering vegetation (∆ED_BPV ≈ 100 to 200) and of herbaceous vegetation cover (∆ED_HDV ≈ 100 to 200) and the reduction of herbaceous vegetation cover (∆PLAND_HDV ≈ less to -10). Concerning the increase of invaded patches edge length (∆ED_CAR, Fig. 6c), it tends to occur at increasing edge length of dune vegetation (∆ED_HDV ≈ 100 to 200) and of artificial areas (∆ED_ART ≈ greater than 200), as well as at decreasing cover of sea class (∆AREA_MEAN_SEA ≈ less to -0.25). As observed with invasion cover (∆PLAND_CAR), also the dimension of invaded patches (∆AREA_MN_CAR, Fig. 6d) tend to increase in correspondence with the reduction of sea-class surface (∆AREA_MEAN_SEA ≈ less to -0.25), with increasing or decreasing urban cover (0.30 < ∆AREA_MN_ART < -0.30) and stable herbaceous vegetation edge length (ED_HDV ≈ 0).

Partial dependence plots for less important change variables in EXP_CAR RF models are provided in Suppl. material 1: figs S12–S15.

In coastal tracts of IAP reduction (RED_CAR), the landscape change variables (RF_∆LM) that explain Carpobrotus spp. contraction (RED_∆LM_CAR; Fig. 7) include both: class metrics of the main cover classes (e.g. ∆LM_HDV, ∆LM_ART, ∆LM_SEA and ∆LM_BPV) and landscape metrics (∆SIDI). Stronger contraction of Carpobrotus spp class cover (∆PLAND_CAR; Fig. 7a) occurs in correspondence with changing beach pioneer vegetation edges length (decreasing ∆ED_BPV ≈ -200 to -300 or increasing ∆ED_BPV ≈ 100), with increasing herbaceous vegetation cover (∆PLAND_HDV ≈ 10), as well as with a consistent increase in artificial areas edges (∆ED_ART ≈ 100 to 200). We also registered a decrease in the number of invaded patches (∆PD_CAR, Fig. 7b) in correspondence with decreasing class metrics of herbaceous vegetation as edge length (∆ED_HDV ≈ -200) and number of patches (∆PD_HDV ≈ -100 to -200) and the increment of urban class cover (AREA_MN_ART > 0). Regarding the decrease in invaded patch edge length (∆ED_CAR, Fig. 7c), it coincides with the reduction in herbaceous vegetation cover (∆PLAND_HDV ≈ -10) and edge density (∆ED_HDV ≈ -200 to -250) and with the simplification of landscape diversity (∆SIDI ≈ -0.025). A consistent reduction on Carpobrotus spp. class patch size (∆AREA_MN_CAR) seems related with increasing values of SEA class metrics as wider surface (∆AREA_MN_SEA > 0.25) and stable edges (∆ED_SEA ≈ 0), as well as with the reduction of artificial surfaces (∆AREA_MN_ART < 0). The dependence trends of other change variables in RED_CAR areas are reported in Suppl. material 1: figs S16–S19 for RED_∆LM_CAR models.

Carpobrotus spp. tends to expand in coastal zones characterised by stable accreting seashore (where the SEA class area remains stable or diminishes) and by increasing urban surfaces (where ART class patches increase). The expansion of Carpobrotus spp. on coastal tracts experiencing seashore accretion and stability may be likely related with the fact that these seashore processes promote the development of herbaceous dune habitats (Bazzichetto et al. 2020) which are known to be especially susceptible for colonisation by invasive alien plants (IAPs) like Carpobrotus spp. (Carranza et al. 2011). Indeed, Carpobrotus spp. primarily expands by displacing herbaceous natural vegetation, confirming the high vulnerability of this natural habitat to IAP invasions. Our results gave new bi-temporal evidence of the habitat preference of Carpobrotus spp. to herbaceous habitats postulated in the past based on static data (e.g. Carranza et al. (2011); Bazzichetto et al. (2018a)) or using a diachronic analysis (Sperandii et al. 2018). The observed increase of IAP’s and urban-patches cover, is likely attributable to the role of artificial areas as a source of non-native species (Carranza et al. 2010) and provides bi-temporal evidence in support of the propagule pressure theory.

The trajectory analysis evidenced a significant rise of all the considered Carpobrotus spp. spatial metrics denoting a consistent process of invasion. Indeed, the extension of invaded areas increased (greater PLAND_CAR) and Carpobrotus spp. tended to be distributed in more numerous (PD_CAR), larger (AREA_MN_CAR) and irregularly-shaped patches (ED_CAR). As observed in other colonisation processes such as forest regrowth (Malavasi et al. 2018b), the increase in the number of patches may denote the emergence of a new nucleus of colonisation, while the enlargement of patch area may indicate the growth of already established invaded points. On the other hand, the observed increase in edge length may indicate the maturity of invasion, with Carpobrotus spp. patches adopting the typical long-shaped pattern along the seashore of the invaded natural herbaceous dune vegetation (Carranza et al. 2010).

The pattern of expansion of Carpobrotus spp. (∆LM_CAR) is significantly associated with landscape dynamics (∆LM_land). Amongst the landscape change variables that best explain Carpobrotus spp. expansion, the seashore stability or accretion (∆SEA close or higher than 0), the increasing surface and edges of urban areas (∆ART) and the size and shape complexity of herbaceous dune vegetation, emerge as particularly influential factors. The observed increase in Carpobrotus spp. cover, patch size and edge length in stable or accreting coastal dunes (e.g. sea class surface reduction), highlights the strong correlation between the presence of dunes, their stability and the heightened susceptibility of landscapes to Carpobrotus spp. invasion. Our bi-temporal analysis-based results provide additional evidence supporting the vulnerability of dunes, a principle previously suggested by invasion risk models using the distance to the shoreline as a surrogate of coastal dune zonation (Bazzichetto et al. 2018a, b).

In tracts registering IAP expansion, Carpobrotus spp. patches become larger and more irregularly shaped in correspondence with an increase in the cover and edge length of artificial areas. This is likely linked to the role of built-up and urban structures, as well as artificial edges in driving invasion processes (Malavasi et al. 2014; Bazzichetto et al. 2018a, b). In these tracts, urban areas may play a double role: an important source of invasive alien propagules and key ecological corridors assuring landscape connectivity for the colonisation of new areas (Boscutti et al. 2022; Lozano et al. 2023). Our results also evidenced that altered coastal landscapes (e.g. detrimentally changed dune vegetation composition and configuration features) might undergo further modification due to the colonisation and spread Carpobrotus spp. (Malavasi et al. 2014). The creation of new artificial corridors (e.g. infrastructures) on fragmented coastal landscapes could aid the invasion process with detrimental effects on dune integrity and natural habitats. As observed for other IAPs, the registered landscape trends may be compatible with further growth of Carpobrotus spp. that conforming dense monospecific carpets could alter coastal dune landscape composition and configuration (Kozhoridze et al. 2022). The colonisation and expansion trends pinpoint the need for planning and implementing dedicated measures to contain and prevent further Carpobrotus spp. expansion at the expense of herbaceous dune vegetation classes which is urgent as it includes several habitats of European conservation concern (EU-2110: embryonic shifting dunes, EU-2120: shifting dunes along the shoreline with Ammophila arenaria; EU-2210: Crucianellion maritimae fixed-beach dunes, EU-2230: Malcolmietalia dune grasslands). These measures encompass both reducing and mitigating dune fragmentation and degradation processes, as well as monitoring activities in the most susceptible landscape elements to aid the implementation of effective prevention and early warning actions.

Within the coastal tracts of Carpobrotus spp. contraction (RED_CAR), landscape spatial-temporal characteristics resulted in being quite dynamic with an intense seashore erosion (SEA class area increase) that constrained the coastal dune zonation to small areas and that curtailed the spatial complexity of overall the natural mosaic (Doody 2004, 2013). In such areas, the SEA category is replaced by beach pioneer vegetation (BPV), that, in turn, substitutes herbaceous dune vegetation (HDV).

The decline in Carpobrotus spp. cover on coastal tracts experiencing seashore erosion may be attributed to the erosion’s detrimental impact on habitats suitable for the invasive plant’s colonisation, such as beach and herbaceous vegetation (Carranza et al. 2011; Bazzichetto et al. 2018a). While this hypothesis has primarily been examined in the context of management activities aimed at eradicating Carpobrotus spp., further research is needed to validate its broader applicability (Chenot et al. 2018).

In these coastal tracts, Carpobrotus spp. tends to be substituted by herbaceous dune vegetation and its pattern in T1 resulted in being simplified into smaller and regularly-shaped patches with respect to IAP pattern on the T0. As evinced by trajectory analysis, the temporal reduction of invaded areas (lower PLAND_CAR) with Carpobrotus spp patches distributed on smaller (AREA_MN_CAR) and regularly-shaped patches (ED_CAR) suggest that Carpobrotus spp. is undergoing fragmentation. As observed in other fragmentation process (Wang et al. 2014; Carranza et al. 2015), the reduction of Carpobrotus spp. area into patches may denote the retreat and disappearance of invaded areas, while the reduction of patch size may indicate the contraction of the remnant invaded points.

The reduction pattern of Carpobrotus spp. (∆LM_CAR) is linked to landscape changes (∆LM_land), specifically those concerning coastal erosion (SEA) and land take (ART). Both processes contribute to the “squeezing” of dune zonation, compressing HDV and BPV communities into simplified small relict areas (Martínez et al. 2014; Gilby et al. 2021). In these tracts, the coastal “squeeze” process reduces suitable land space for coastal dune ecosystems and, consequently, the possibility to maintain their essential functions (Martínez et al. 2014). Under these “squeeze” conditions, only hard structures, such as building and human infrastructures, remain, while both Carpobrotus spp. patches and native vegetation tend to diminish and eventually disappear.

Within one decade, Carpobrotus spp. registered a decline on cover, patch size and edge length which occurred together with the simplification of natural and semi-natural land-cover classes (e.g. the reduction on cover, edge length and aggregation of herbaceous and pioneer dune vegetation) and overall landscape diversity (SIDI). Moreover, this reduction coincides with increased artificialisation (expanding urban areas in terms of area and edge length) and seashore erosion (expansion of sea area and edge length). The widespread decrease in landscape diversity and the deterioration of dune integrity (Acosta et al. 2003; Drius et al. 2013), combined with the fragmentation (sensu Wang et al. (2014)) observed in the invasion pattern of Carpobrotus spp., suggest a substantial influence of both abiotic factors (seashore erosion) and human-driven forces (urbanisation) on shaping the entirety of the coastal dune mosaic, encompassing both natural and invaded areas.

The authors have declared that no competing interests exist.

No ethical statement was reported.

This work was supported by the bilateral programme Italy–Israel DERESEMII-CC (developing state-of-the-art remote sensing tools for monitoring the impact of invasive plant species in coastal ecosystems in Israel and Italy under climate change) funded by the Italian ministry of foreign affairs and international cooperation and the israel ministry of science and Technology. It was also partially supported by the National Recovery and Resilience Plan (NRRP), Mission 4 Component 2 Investment 1.4 – Call for tender No. 3138 of 16 December 2021, rectified by Decree n.3175 of 18 December 2021 of the Italian Ministry of University and Research funded by the European Union – NextGenerationEU, Project code CN_00000033, Concession Decree No. 1034 of 17 June 2022 adopted by the Italian Ministry of University and Research, CUP J83C22000870007; CUP F83C22000730006; CUP H73C22000300001; Project title “National Biodiversity Future Center – NBFC”. The research was also funded by PRIN 2022JBP5F8- PREVALIEN. Enhancing Knowledge on Prevention and Early Detection of the Invasive Alien Plants of (European) Union concern in the Italian Protected Areas. CUP Master: J53D2300657-0006.

Flavio Marzialetti: Conceptualisation, Data curation, Formal analysis, Validation, Investigation, Methodology, Visualisation, Writing – original draft, Writing – review and editing. Giacomo Grosso: Data curation, Formal analysis, Validation, Investigation, Visualisation, Writing – original draft, Writing – review and editing. Alicia Teresa Rosario Acosta: Conceptualisation, Funding acquisition, Supervision, Writing – original draft, Writing – review and editing. Marco Malavasi: Methodology, Supervision, Writing – original draft, Writing – review and editing. Luigi Cao Pinna: Formal analysis, Validation, Investigation, Methodology, Writing – original draft, Writing – review and editing. Marcelo Sternberg: Supervision, Funding acquisition, Writing – original draft, Writing – review and editing. Sharad Kumar Gupta: Methodology, Writing – original draft, Writing – review and editing. Giuseppe Brundu: Supervision, Writing – original draft, Writing – review and editing. Maria Laura Carranza: Conceptualisation, Supervision, Funding acquisition, Investigation, Visualisation, Methodology, Writing – original draft, Writing – review and editing.

Flavio Marzialetti https://orcid.org/0000-0001-5661-4683

Alicia Teresa Rosario Acosta https://orcid.org/0000-0001-6572-3187

Marco Malavasi https://orcid.org/0000-0002-9639-1784

Luigi Cao Pinna https://orcid.org/0000-0002-1152-258X

Marcelo Sternberg https://orcid.org/0000-0001-8710-4141

Sharad Kumar Gupta https://orcid.org/0000-0003-3444-1333

Giuseppe Brundu https://orcid.org/0000-0003-3076-4098

Maria Laura Carranza https://orcid.org/0000-0001-5753-890X

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