This study was carried out in coastal dune areas along the Channel and the North Sea, covering the North of France, Belgium, the United Kingdom and the Netherlands (Fig. 1). Within this area, we focussed on sandy coasts with marram-dominated, yellow dunes. This area included the location of S. inaequidens settlement and the northernmost location within its distribution in coastal dunes, thus enabling us to study the front of the ongoing invasion.

Figure 1.

The samples included in the analysis. Colours indicate the different countries. Senecio inaequidens was not found in the UK. Map made with QGIS v3.6 (QGIS Development Team 2021).

For a recent biodiversity study, 46 dune transects spread along the study area were selected. The transects had a mean length of 1212 m (shortest: 230 m, longest: 3348 m) and were located within the first 100 m from the front of the foredunes. Within each transect a number of sampling locations was chosen based on the length of the dune transect with an average of 14 samples (min 5; max 37). Each sampling location was characterized by a central marram grass tussock. Individual sampling locations were separated by at least 20 m and chosen with the aim to maximise the variety of surrounding marram grass configurations. For the total number of samples and transects per country, see Table 1.

The occurrence of narrow-leaved ragwort (Senecio inaequidens) was mapped at each sampling location. The number of S. inaequidens plants was counted within a radius of 5 m around the central marram grass tussock for those sampled in France, the UK and the Netherlands. Due to a change in the protocol of the biodiversity study, in Belgium the occurrence was scored into four categories: “not present”, “sparse”, “moderate” and “abundant”. Data on the occurrence of S. inaequidens were collected during three consecutive summers: in July 2017 data were collected along the Belgian coast; in July, August and September 2018 along the French coast; in August and September 2018 and June 2019 along the Dutch coast; and in July and August 2019 along the coast of the UK (Norfolk and Devon).

From available vegetation maps of the foredunes (Bonte et al. 2021), the proportional cover by marram grass in the vicinity of the central marram grass tussock (P), together with a measure of spatial autocorrelation of marram grass occurrence (normalised join count statistics, JC; Cliff and Ord 1981), were calculated. These two parameters were used to express the spatial configuration of the surrounding marram grass. The proportion of marram grass cover is straightforward and ranges from 0 (no marram grass present) to 1 (the whole area is covered with marram). The measure of spatial autocorrelation is negative when the marram grass is regularly distributed in the landscape. If the marram grass is randomly distributed, the parameter is close to 0 and it is positive when the marram grass occurs clustered together. As pointed out by Bonte et al. (2021), marram grass distribution is almost always clustered (i.e., high JC values) and rarely random. These two parameters (P and JC) were calculated within four circles with different radii (5 m, 10 m, 20 m, 50 m) around the central marram grass tussock to represent different spatial scales. See supplementary material of Bonte et al. (2021) for a more in-depth explanation of construction of the vegetation maps and calculations of the spatial parameters.

To study the effect of S. inaequidens on marram grass growth, we performed a growth experiment with a split-plot design: sand affected by S. inaequidens was gathered at the Belgian coast together with bare sand for the control group. Half of the volume of sand gathered was sterilised (by autoclaving at 121 °C/1 bar for 30 minutes) in both groups to determine whether any observed effect could be biotic or abiotic.

Sand was gathered from three different sites situated on the western, central and eastern Belgian coast: in the foredunes in Oostduinkerke (Ter Yde) for the west coast, for the mid coast in Oostende (Fort Napoleon) and for the east coast between Wenduine and Zeebrugge (two locations were used due to low occurrence of S. inaequidens). Ten plots were sampled at each site (for a total of 30 plots). Each plot yielded two samples: 2L rhizospheric sand from underneath S. inaequidens plants and 2L of bare sand taken 5–10 m away. This way, changes in soil between two paired samples, other than due to the influence of S. inaequidens, were minimised. The sand was stored in the fridge (max 3 days) to assure the survival of the soil biota until the sand was used. The 2L samples were divided into two 1L sub-samples from which one was sterilised and the other was not. Thus, we had four treatments: Senecio-influenced vs. bare sand at the plot level combined with sterile vs. non-sterile soil at the subplot level (Fig. 2). In other words, influence of Senecio was the whole-plot factor and soil sterilisation the subplot factor, with whole plots organized in pairs, which act as statistical blocks.

Figure 2.

Split-plot design of the growth experiment for the site at Oostduinkerke (Western Belgian coast). 2L samples of sand, (1) sand from unvegetated locations or (2) sand from underneath Senecio, were split into two 1L subsamples, one of which was sterilized. This enabled us to investigate whether the effect of Senecio was achieved via the biotic or abiotic portion of the soil. Map made with QGIS v3.6 (QGIS Development Team 2021). Aerial photograph (summer 2018) source: Agency for Information Flanders (geopunt.be).

Marram grass seedlings were used for the experiment because seedlings are more susceptible to environmental influences than fully grown plants (Huiskes 1979). The seedlings were grown from seeds gathered at the Belgian coast (Oostduinkerke, ter Yde) from the same population in order to minimize genetic effects. The seeds were collected during the summer of 2019 and stored at room temperature in the lab. All seeds were surface-sterilised as in de la Peña et al. (2010) before they were left to germinate under standardized conditions (on commercially available sand saturated with demineralized water; photoperiod: 16/8 h light/dark; temperature: 22 ± 1 °C) for 2 weeks prior to dune sand collection. As a baseline, the whole seedlings were weighed and the length of roots and leaves was measured before planting.

All 120 pots (3 sites × 4 treatment combinations × 10 plots) were filled with 1L of sand in which three seedlings were planted. The pots were placed in a growing chamber under the same conditions as mentioned before for the germination of the seeds. All pots were watered twice a week, on the same day, with demineralised water until near-saturation. Each pot was labelled with a unique ID in order to prevent observer bias.

After 2 weeks of growing, the largest seedling was selected to grow for another 10 weeks. The other two seedlings were removed. This was done to ensure that all remaining seedlings had rooted properly in order to minimise die-off and resulted in only three plants dying during the whole experiment (one from each treatment, except for the sterilized bare sand treatment). At the end of the growing period the whole plants were collected, all leaves were counted and the length of the longest leaf and root was measured. Further, all leaves and roots were weighed separately, both before and after drying in an oven at 70 °C for 48 h.

Senecio inaequidens was observed at 176 of the 408 sites included in the analysis. The most northern and southern transect where S. inaequidens was observed are respectively at Wassenaar (52.1565°N, 4.3404°E; the Netherlands) and Wimereux (50.7931°N, 1.6074°E; France). S. inaequidens was most frequently present in Belgian samples, followed by France and the Netherlands (Fig. 3). In the United Kingdom, no S. inaequidens was observed.

Figure 3.

The average occurrence of S. inaequidens, calculated as the proportion of samples within each transect where S. inaequidens was found. BE = Belgium; FR = France, NL = the Netherlands.

The four models selected were all at the 5 m scale (using an AICc delta value of 2; see Table 2 for the selected models; see Suppl. material 1: Table S1 for all models), implying that S. inaequidens reacts to marram grass spatial configuration at small distances. The predicted occurrence is depicted in Fig. 4a. When the vegetation is highly clustered together (high JC values), the occurrence of S. inaequidens is negatively correlated with marram grass cover. Further, we see a clear minimum probability of occurrence of S. inaequidens at more random distributions of marram grass (low JC values) with intermediate vegetation cover. Since no random configurations were found at intermediate P, the minimum is probably due to a small number of samples with a low JC (see Fig. 4b), thus caution is advised when interpreting this result. Although we actively tried to sample in a wide range of different spatial configurations of marram grass, there is still a low number of data points with low marram grass cover and less clustered configurations due to the nature of the system. To ascertain that the outcome of the analysis was not greatly influenced by those few samples, the analysis was redone after excluding those samples. The results did not differ greatly (see Suppl. material 1: Fig. S2).

Figure 4.

a The overall relation between the probability of occurrence of S. inaequidens and the spatial configuration of marram grass. The colours indicate the probability of occurrence as %. b Density distribution plots of the observed cover (P) and spatial autocorrelation (JC) of marram grass within a 5 m radius of the central marram grass tussock. This plot only contains the data of the transects where S. inaequidens was found. Colours indicate whether S. inaequidens was present (yellow) or absent (purple).

The first PC of the PCA of all measured plant traits explained 73.9% of the variation, while the second PC explained 14.7%. Scores along PC1 were significantly correlated with all plant traits (see Suppl. material 1: Fig. S1, Table S2), therefore, we used PC1 as a reliable indicator of overall plant growth. Marram grass growth was affected by both the abiotic and biotic components of the soil (Fig. 5). However, no significant interaction was found (F_1,82.7 = 1.10, p = 0.298). Soil sterilisation had a positive effect on marram growth (F_1,83.4 = 106, p < 0.001), which means that the soil biota had a negative effect on marram biomass. The plants grown on sand from underneath S. inaequidens grew better than plants grown on bare sand (F_1,82.7 = 59.2, p < 0.001).

Figure 5.

Box- and violin plots represent distribution of PC1 values for marram grass growth. Horizontal lines above the boxplots indicate comparisons between treatments, *** indicate significant difference of p < 0.001. Colours indicate whether biota were present (yellow) or absent (purple). Number of samples per treatment is 29, except for sterilized sand from unvegetated locations, where it is 30.

No evidence was found for the hypothesized optimum probability of establishment of S. inaequidens at intermediate marram grass densities. In fact, our results indicated that S. inaequidens has no problem growing in sandy conditions, as we observed a negative correlation between vegetation cover and probability of establishment. This indicates that S. inaequidens is more susceptible to competition than to sand burial. Indeed, some studies found that this species is a good coloniser rather than a good competitor (Scherber et al. 2003; Caño et al. 2007; Thébault et al. 2011). Furthermore, the available area to root in – in this case open sand – also decreases as vegetation cover increases. This is especially important since S. inaequidens is an annual plant which relies on high propagule pressure to spread (Thébault et al. 2011).

Due to the nature of the system, higher proportions of marram grass occur mainly towards later stages of succession. In these later stages, marram starts to decay and the spatial configuration starts to return to a more random distribution (i.e. lower JC values and slightly lower P values) because marram grass is slowly being replaced by other plant species. This leads to a rise of the probability of Senecio establishing which may indicate that it is becoming a stronger competitor.

Overall, the probability of establishment of S. inaequidens displays high values across the whole range of sampled natural marram grass configurations. Since we aimed to maximise the variety of natural marram grass configurations surrounding the sample, configurations that were not sampled probably do not, or not often, occur in nature. In fact, such configurations arise probably mainly when marram grass is planted (i.e., for coastal protection) and afterwards when the planted dune is ‘maturing’. This makes it hard to extrapolate our findings to these specific situations.

We hypothesised that the effect of S. inaequidens on marram grass growth would be negative, mainly because of interactions with the soil community. However, we concluded that the overall effect is positive. This effect is purely abiotic, since there is no significant interaction between sand sterilisation treatment and the provenance of the sand (underneath/away from Senecio). Similarly, intraspecific plant-soil feedbacks from Senecio jacobaea are also known to be (partly) abiotic (Wang et al. 2019), although the effect was negative in the cited study. Dassonville et al. (2008) showed that invasive species (S. inaequidens being one of the species included in their review) can have a positive effect on nutrient concentrations in plots with initially low concentrations. Since sand indeed has low nutrient concentrations (Reijers et al. 2020), this explanation is applicable here. Currently, we cannot verify this explanation since nutrient concentrations were not measured in the experiment.

Because marram grass growth was promoted in sand influenced by S. inaequidens, we can conclude that pyrrolizidine alkaloid concentrations had no, or a negligible, negative effect on marram grass. This is not surprising, since the most probable mechanism of PA enrichment of the soil is via passive release from roots and leaf litter (Joosten and van Veen 2012; Nowak et al. 2016; Selmar et al. 2019) and because some plants are even known to take up PAs without experiencing adverse effects (Nowak et al. 2016; Selmar et al. 2019). In contrast, Ahmed and Wardle (1994) found a negative effect of PA on plant growth. In our study, such a negative effect of PAs may have been counteracted by the simultaneous nutrient enrichment caused by S. inaequidens. This is in line with the observations by Reijers et al. (2020) that marram grass is more capable of coping with stressful conditions when nutrient availability is higher.

We observed a significant negative effect of soil biota on marram growth, with sterilisation of the soil having a positive effect on the biomass of marram, independent of the sand origin. This indicates that soil biota in the Senecio rhizosphere have approximately the same (negative) effect as the community within unvegetated sand. Thoden et al. (2009) found that PAs suppress the development of juvenile Meloidogyne hapla nematodes. Species from this genus also colonise marram grass. However, it was suggested before that nematodes from this genus do not develop to adults on marram grass anyway (Van der Stoel et al. 2002). On the other hand, Pratylenchus nematodes are able to colonise both Senecio (Zasada et al. 2017) and marram grass roots (Van der Stoel et al. 2002), which would enable infection of marram roots by Pratylenchus spp. present on Senecio roots. However, arbuscular mycorrhizal (AM) fungi have low levels of endemism and host specificity (Davison et al. 2015; Aldorfová et al. 2020) and are known to colonise roots of Senecio spp. (van de Voorde et al. 2010; Alguacil et al. 2012; Reidinger et al. 2012). Since both European and American marram grass (Calamagrostis arenaria and C. breviligulata resp.) benefit from AM fungi when faced with nematode infection (Little and Maun 1996; De La Peña et al. 2006), it is possible that the negative effect of the nematodes is counteracted by the AM fungi also accrued on the Senecio roots. Furthermore, some studies have shown that Senecio species effectively reduce density or diversity, depending on the study, of whole soil communities (Kowalchuk et al. 2006; Thébault et al. 2010; Harkes et al. 2017). However, since these studies focussed on the effect on whole soil communities, it is difficult to make predictions for marram grass specifically. We conclude that PA concentrations in the field did not reduce the overall negative effect of the soil community, either because the community as a whole was affected and thus both negative and positive elements therein were reduced or because PA concentrations are too low to affect the soil community in general.

Our results indicate that the biotic soil community surrounding Senecio roots has approximately the same (negative) effect as the community within sand without plants growing in it (i.e., no significant interaction effect). Since endoparasites are known to be more damaging to marram grass (Van der Putten and Van Der Stoel 1998), it could be that we excluded their effect because we did not use an inoculum from Senecio to infest the soil and consequently marram roots with endoparasites. On the other hand, the biota accumulating on the senecio roots could not be compatible with marram grass roots at all. Consequently, sand from unvegetated patches, which is thought to function as a temporary ‘enemy-free’ space for marram grass to root in (Van der Stoel et al. 2002), has the same biological effect on marram growth as Senecio-influenced sand.

The observed positive effect of sterilisation in the unvegetated sand is caused by soil biota, such as nematodes, who have survival stages that can disperse in the dunes (e.g., Heterodera cysts) and subsequently colonise the marram grass roots in the lab (e.g., De Rooij‐Van Der Goes 1995; Van der Stoel et al. 2002). Indeed, studies investigating marram grass PSFs frequently used sand from the beach or even the sea floor for their control treatment (Van der Putten et al. 1988; Van der Putten and Troelstra 1990). However, we decided to take sand from the foredunes in order to maximise similarity of soil characteristics between Senecio-affected and unvegetated paired samples.

Since we only studied correlations, it could be that S. inaequidens established only on the more nutrient-rich sand in the dunes, which would in turn explain why marram grass grows better in this sand. However, this is very unlikely since dunes are extremely dynamic and hence the top layers of sand are thoroughly mixed, creating a homogenously resource-poor environment (Reijers et al. 2020). Furthermore, marram grass was growing in the dunes long before Senecio, thus, if there would be patches with more nutrients, those patches would likely already have been occupied by marram grass. When sand is fixated by plant roots, the mixing is halted and nutrient heterogeneity can start to occur. Since we took sand from the rhizosphere of S. inaequidens, higher nutrient availability caused by S. inaequidens is still a viable explanation for our findings.

Sandy habitats, such as coastal dunes, are characterised by unstable substrate with many open patches of bare sand in between the vegetation. These patches are ideal opportunities for the establishment of new species (Axmanová et al. 2021). From our field survey, we can conclude that S. inaequidens is indeed capable of colonising these open patches. The results from our lab experiment further indicate that establishment of S. inaequidens can enhance marram growth in particular, but probably also plant growth in general, after it dies off and nutrients become homogenised. Analogous positive effects on growth of co-occurring plants have been found for other Senecio species (van de Voorde et al. 2011). This enhanced plant growth can lead to an overall shift towards more vegetated dunes, thus further intensifying the worldwide trend towards dune stabilisation (Gao et al. 2020). Dune stabilisation directly implies lower sediment transport to dune parts further inland, which may enable establishment of other species and hence accelerate natural succession.

For marram grass specifically, reduced sediment supply due to dune stabilisation leads to a shift towards a more clustered vegetation configuration (Reijers et al. 2021) which optimises sand capture at small spatial scales (Reijers et al. 2019). In contrast, the potential for dune formation at larger spatial scales will be reduced, affecting dune geomorphology as a whole, ultimately resulting in lower dunes (Reijers et al. 2021). If lowered sediment supply indeed also accelerates dune succession, marram grass will be replaced more rapidly by other plant species less capable of forming dunes.

In conclusion, invasion of dune ecosystems by S. inaequidens could lead to a shift in sand dynamics by colonising bare sand patches, in turn accelerating the natural succession of dune vegetation. This could hamper dune growth and further reduce dune height. A reduction in dune height could in turn compromise coastal protection, since higher dunes are known to better protect the hinterland (Zarnetske et al. 2012; Seabloom et al. 2013).