In 1952 and 1953, 68 permanent plots were installed in the northern part of Meijendel (Fig. 1) as part of the Dutch 'Network Ecological 'Monitoring' (Smits et al. 2002). Their sizes range between 4 m² and 225 m², related to the suggested plot sizes for different vegetation types (Schaminée et al. 1995a). The plots are situated in three parts of Meijendel. The western cluster is in the zone of the parabolic dunes (Helmduinen); the other two are situated in the large dune valleys of Kijfhoek and Bierlap (Van der Meulen et al. 1985). The plots cover all successional stages: dune grasslands (Phleo-Tortuletum ruraliformis, Taraxaco-Galietum veri, Festuco-Galietum veri), dune shrubs (Hippophao-Ligustretum), and tall shrubland and forests (Alno-Padion communities, Rhamno-Crataegetum, Crataego-Betuletum). In the parabolic dunes, also plots were laid out comprising two rare vegetation types: (i) communities of the Anthyllido-Silenetum (Schaminée et al. 1998) and (ii) communities of the Caricion davallianae (Schaminée et al. 1995b). All permanent plots are subject to livestock grazing; thus, we could not apply a comparison with ungrazed plots.

Figure 1. 

Distribution of the selected permanent plots in Helmduinen, Kijfhoek and Bierlap in Meijendel. Upper left corner: North Sea, right side: inner dune edge near Wassenaar. Blue areas: infiltration ponds for artificial recharge, light blue: dune slacks. Bottom righthand corner: cut-out of Meijendel, surrounded by the cities of The Hague and Wassenaar.

We investigate the vegetation changes in the dry part of the dune ecosystem (see Schaminée et al. 1996; 1998), covering more than 95% of the total area. We excluded the permanent plots in the moist and wet parts as they lie outside the study's scope (Van der Hagen 2022). We evaluate 41 permanent plots (60% of the available plots), which over time sum up to 479 relevés. In 1953, almost all dry dune ecosystem vegetation types were present (Appendix I).

Following the internal homogeneity principle, the plots initially consisted of a homogeneous vegetation patch reflecting a single syntaxon (Braun Blanquet 1951, 1979). Over time, they developed heterogeneously into two or more vegetation types. Initially, the permanent plots were monitored yearly using the Braun-Banquet approach (Braun-Blanquet 1951; Schaminée et al. 1995a). Since the 1970s, the recording cycle decreased to about once in every four years and different scales were applied to record species abundances (e.g., Braun-Blanquet; Barkman, Doing & Segal; Londo; see Schaminée et al. 1995a). The varying scales were transformed to an ordinal scale (0-9) using the standard procedure in the database management system of TURBOVEG v2.3 (Hennekens and Schaminée 2001; Hill and Šmilauer 2005).

The syntaxonomic nomenclature follows Schaminée et al. (1996; 1998; 2010), Stortelder et al. (1999), and Kopecký and Hejný (1974). The taxonomy of vascular plants is according to Van der Meijden (2005), mosses according to Siebel and During (2006), and lichens according to Van Herk and Aptroot (2004).

Lammers (2015) applied the TWINSPAN algorithm (Hill and Šmilauer 2005) and ASSOCIA (Van Tongeren et al. 2008) in the TURBOVEG v2.3 program to classify the plot-based dataset into clusters. These classification tools were used to identify the various successional stages, ranging from bare sand with open grassland to closed forest. To allocate found vegetation clusters to syntaxonomic names and to define groups of characteristic species, Schaminée et al. (1995a; 1996; 1998) and Stortelder et al. (1999) were applied. Lammers (2015) named the distinguished vegetation clusters according to the defined syntaxonomic species groups (Table 1). To analyse vegetation changes over time, the names of identified plant communities at each point in time were applied. For years in which a specific plot was not surveyed, the name of the previously identified vegetation cluster continued.

In 1953, seven out of nine successional stages were present with a dominance of tall shrubland of Rhamno-Crataegetum with forest elements (8), Taraxaco-Galietum, a dune grassland of lime-rich soils (2), and the pioneer grassland Phleo-Tortuletum (1) (Fig. 2; Appendix I). The rare and species-rich Anthyllido-Silenetum (3) was also present in some plots. Two successional stages were absent: the Festuca-Agrostis basal community characterized by acidic topsoils (4), and the Quercus robur forest stands with tall shrubland elements (9) (Fig. 2).

In 2012, only five out of nine successional stages were still present in the permanent plots, including early and late successional stages (Fig. 2). In comparison to 1953, dune grasslands Phleo-Tortuletum and Taraxaco-Galietum (1, 2) had a limited presence in 2012. On the other hand, the Quercus robur stands with tall shrubland elements (9) were already present since 2004 (Appendix I) and showed up in an increasing number of plots up to 2012. Over time, the Rhamno-Crataegetum with elements of low shrubs (7) and the tall shrubland of Rhamno-Crataegetum with forest elements (8) became more prominent with a peak in 1992 (Fig. 2). The successional stages 3, 4, 5, and 6, including the rare Anthyllido-Silenetum, disappeared in all plots in 2004 (Appendix I). Overall, changes observed since 1990 predominantly led towards both extremes of successional stages, with a strong emphasis on the late successional stages (Fig. 2; Appendix I).

In the pre-1990 period, when no livestock grazing occurred, we observed a balance in changes between progressive and regressive succession. Since 1990, progressive succession has dominated after the re-introduction of livestock grazing (Table 3). More precisely, from 1953 until 1970/1972, progressive and regressive changes occurred between all successional stages. After 1970/1972, we observed a slight shift towards progressive succession, with an emphasis on tall shrubland of the Rhamno-Crataegetum and, after 1995, also into closed Quercus robur forest stands with tall shrubland (9; Appendix I). Both vegetation types became increasingly prominent from 2004 onwards.

The lime-rich dune grasslands of the Taraxaco-Galietum (2), substantially present in 1953, decreased after 1972 and again more pronounced after 1981. After 1988, lime-rich grasslands almost completely vanished (Fig. 2; Appendix I). The Taraxaco-Galietum returned from 2004 to 2012 in a limited number of plots (Appendix I). Most of the lime-rich dune grasslands (1, 2) shifted towards the Rhamno-Crataegetum with grassland elements (6) or with elements of low shrubs (7) (Appendix I). In 1981, grasslands of the Festuca-Agrostis basal community (4) with acidic topsoil appeared.

In the post-1990 period with livestock grazing, only a few regressive shifts occurred (10%: Table 3). Post-1990, the Rhamno-Crataegetum communities (7, 8) became very prominent. Since 2004, Quercus robur forests (9) make up a substantial part of the permanent plots (Appendix I). In 1998, the Festuca-Agrostis basal community (4) partly transitioned into lime-rich grasslands (1, 2). Soon after the introduction of livestock in 1990, the sandy pioneer grasslands of the Phleo-Tortuletum (1) disappeared but re-occurred after 2004 (Fig. 2; Appendix I). The rare Anthyllido-Silenetum (3) persisted from 1953 to 2004, even slightly increasing from 1958 to 1973. In 2004, however, it disappeared in all plots.

Three driving factors may explain succession in coastal dune grassland habitats in coastal dunes of North-Western Europe: the almost absence of rabbit grazing and digging, the autonomous return of blowouts after 2001, and the decrease of nitrogen deposition after 1990 (Van der Hagen et al. 2020a; 2020b; Fig. 3). Like livestock, rabbits also feed on grassland vegetation and can significantly reduce above-ground phytomass when occurring in high numbers (Drees 1988). The sharp decline of rabbits due to viral diseases is likely to be the most decisive due to their preference for young nutrient-rich shrub and tree seedlings (Drees and Olff 2001; Van Tongeren 2006; Van der Hagen et al. 2020b). Additionally to the grazing patterns, the burrow digging by rabbits also affects the vegetation composition and structure. The sand taken outside of the burrow creates small patches of bare sand, characteristic for plant communities of the early successional stage of the Phleo-Tortuletum. Furthermore, collapsed burrows give way for blowout development (Pluis 1986; Jungerius and Van der Meulen 1988). Both processes create open vegetation structures and facilitate the germination and growth success of annual and low-competitive plants (Van der Hagen et al. 2020b; Aggenbach et al. 2018).

In 1954 and 1989, the rabbit population in The Netherlands was reduced by 90%-95%, following the disease outbreaks of myxomatosis (Drees and Olff 2001) and the Rabbit Viral Haemorrhagic Disease (RVHD-1; Drees and Van Maanen 2004; Scheffer 2012). After a slight recovery, RVHD-2 reduced the number of rabbits again in 2015 (IJzer et al., 2016). The sharp decline in rabbit numbers due to epidemic viral diseases was a window of opportunity for the establishment and growth of shrub and tree species and the establishment of late successional stages observed in our data.

Up to 1990, we observed an equilibrium of progressive and regressive succession in our study plots (51% progressive shifts versus 49% regressive: Table 3). However, since 1970/1972, a slight shift was detectable (Appendix I) that we link to the fact that it may take some years for a shrub or tree seedling to cover a noticeable area in the plot. Since 1990, when livestock was re-introduced, the observed shifts towards late successional stages were even more pronounced (90% progressive, 10% regressive: Table 3). Shrubs and trees in vegetation types of the Rhamno-Crataegetum with grassland (6), shrubs (7) and shrubs with forest elements (8) became more common up to 2012. In 2012, grassland stages of the successional stages of the Phleo-Tortuletum and Taraxaco-Galietum remained in just a few plots.

Unlike rabbits, livestock are generalist grazers, that do not differentiate between grasses and herbs and do not specifically feed on young seedlings of shrubs and trees (Bokdam 2003; Lamoot et al. 2005; Bakker 2003; Bakker et al. 2009; Smit and Ruifrok 2009; Runhaar et al. 2015). In contrast to the impact of rabbits (Drees 1988; Bakker 2003; McNab n.y.), the livestock grazing pressure of 0.06-0.07 LLU*ha^-1*year^-1 may have been too low to graze most of the seedlings of shrubs and trees in the Meijendel dune area (Ranwell 1960; Olff and Boersma 1998).

Figure 3. 

The Meijendel dune area: a landscape view with encroachment of high shrubs and the reopening of open sandy dunes after 2001 (A) and a vegetated blowout with an open southern and grassland northern exposed slope with only one rabbit burrow left (B). Photo credit: (A) H.G.J.M. van der Hagen, 2008; (B) F. Witte, 2011.

In 1954, the devastating rabbit disease myxomatosis killed most of the rabbits in the Netherlands (Drees and Olff 2001). We did not observe progressive succession before 1970/1972, but we cannot directly attribute the substantial drop in rabbit numbers to our results. However, the effect of reduced grazing and burrow digging by rabbits might be delayed and is likely to be visible after some time. For example, from 1970 to 1972, we observed increasing progressive succession with shifts from the open grasslands (1, 2) and from the Rhamno-Crataegetum with grassland elements (6) towards the Rhamno-Crataegetum with elements of low shrubs (7) (Fig. 2; Appendix I). Similar results on progressive succession following rabbit reduction were found by Van Groenendael et al. (1982), Bakker (2003) and Van Tongeren (2006).

Growth ring counts on Crataegus monogyna revealed a direct link between the high presence of Crataegus shrubs and the first three years after the 1954 myxomatosis outbreak (Salman and Van der Meijden 1985), indicating the imminent influence of rabbits on seedling development. In the low presence of rabbits, seedlings can mature even in the presence of livestock or when germinating close to thorny shrubs or poisonous plants like Jacobaea vulgaris (Vera et al. 2006). Thorny and long-lasting shrubs, like Crataegus monogyna, gradually dominate and nurse seedlings of Quercus robur and Betula species (Vera et al. 2006). These changes indicate the start of a Rhamno-Crataegetum with forest elements that gradually changes into forest stands with Quercus robur.

In our study, shrublands of Hippophae rhamnoides (Table 1: Hippophao-Ligustretum plant community in clusters 6 and 7) increased from 1975 to 1990 (Van der Hagen et al. 2020b). After 1990, large parts of the Hippophae shrubland collapsed and returned to grasslands. This is due to the lifespan of Hippophae of around 40 years, which equals the timespan between the 1954 outbreak of myxomatosis and 1990 (Van der Hagen et al. 2020b) but note that the lifespan of Hippophae can vary between 18 and 56 years, incidentally up to 80 years (Zoon 1995; Provoost and Declerck 2020; Decuyper et al. 2020). However, our permanent plot data showed that shrubland shifted mainly into tall shrubs (e.g., Crataegus monogyna; Van der Hagen 2020b) from 1972 to 2012 (Fig. 2: 6, 7, 8) and later into Quercus robur forest stands (Fig. 2: 9).

In 1989, just before the introduction of livestock, another devastating rabbit disease rVHD-1 killed large parts of the revived rabbit population (Scheffer 2012). As before, we could not see any effect of the substantial reduction of rabbits due to rVHD-1 in our data. However, we observed an apparent increase in late successional stages from 1990 onwards, likely to be explained by the reduced number of rabbits again.

At the end of our series of permanent plots in 2012, dry dune grasslands were almost absent except for the Taraxaco-Galietum that persisted throughout the study period. The most diverse vegetation type of the 'Zeedorpenlandschap', the Anthyllido-Silenetum, vanished from the permanent plots in 2004. This plant community is mostly restricted to the Netherlands and facilitated by multiple aspects of direct and indirect anthropogenic low-impact activities in the dunes, like hunting, small-scale agriculture, and animal husbandry, including livestock grazing (Slings 1994; Van der Hagen 2022). Due to the introduction of livestock, one would expect the return of this grassland stage. Conversely, it disappeared in the Meijendel plots (see also Mourik 2004; Van den Bos 2007; Van der Hagen et al. 2015).

After 2004, we observed a return of the open dune grassland of the Phleo-Tortuletum (1). This increase, however, could not be linked with the 1990 introduced livestock. Several other reasons may be responsible, including the autonomous increase of blowouts since 2001 along the Dutch coast (especially in lime-rich dune habitats like Meijendel), the observed reduction of atmospheric nitrogen deposition (Aggenbach et al. 2018; Van der Hagen et al. 2020b), and climate change with more weather extremes. Heavy rains induce water erosion exposing lime-rich sand vulnerable to wind transport (Jungerius and Van der Meulen 1988).

The permanent plots forming the basis of our study were not randomly distributed when installed in the 1950s. Nevertheless, they covered a meaningful and diverse range of dry grassland communities and as such offered a reliable basis for the observed shifts in Fig. 2. According to Chytrý et al. (2014), utilizing solely permanent plot data for vegetation succession studies can lead to false significant changes and false observation of no changes. Thus, the use of permanent plots might give only limited information on the successional processes occurring in an area over time. However, our study's results align with Van der Hagen et al. (2020b; 2022), proving similar changes in vegetation structure using aerial photographs and with Van der Hagen et al. (2020b) in a since 1975 rabbit exclosure experiment.

Other factors affecting vegetation changes in the dune landscape might interfere with the developments in the permanent plots. For example, nitrogen deposition, peaking around 1990, is likely to have affected progressive succession (Noordijk 2007; Kooijman et al. 2009; Van der Hagen 2020b). Although the lime-rich dune habitat of Meijendel is largely buffered against nitrogen deposition (Aggenbach et al. 2017), nitrogen levels are still high (Bobbink and Hettelingh 2010). Other factors concern climate-change related effects, specifically temperature rise, changes in rainfall and seasonality (KNMI 2013), and the increase of the topsoil temperature, resulting in accelerated soil processes (Bakema et al. 2022).