Accommodation space indicates dune development potential along an urbanized and frequently nourished coastline

With densely populated areas well below mean sea level, the Netherlands relies heavily on its dunes to ensure coastal safety. About half of the sandy coastline, however, is subject to structural marine erosion and requires frequent sand nourishments as a counteractive measure. A key component of present-day coastal safety policy is creating favorable conditions for natural dune development. These conditions essentially involve a (1) steady supply of wind-blown sand towards (2) wide accommodation space where sand can accumulate and dunes are sheltered from frequent storm surge impact. This paper 5 examines to what extent an experimental mega-scale beach nourishment (termed Zandmotor in Dutch) has contributed to creating accommodation space favorable for dune development. Using publicly available airborne Lidar data and Sentinel-2 satellite imagery, favorable accommodation space is identified by comparing recent changes in coastal morphology against dune vegetation cover dynamics. With a focus on European marram grass (Ammophila arenaria) as the most prominent dunebuilding species, this paper demonstrates that the Zandmotor supports an especially high potential for incipient (embryo) dunes 10 to develop as most of its favorable accommodation space is located on the beach. However, considering the conditions required for successful marram grass establishment as well as persistent anthropogenic disturbances arising from recreation and nature management practices, it is not likely that dune development along this urbanized coastline reaches its full potential.

quality and natural values, outweigh the extra costs involved and to determine to what extent such approach can help cope with expected changes in the global climate (e.g., Mulder and Tonnon, 2011;Stive et al., 2013;De Schipper et al., 2016). In line with the building-with-nature approach, natural dynamics are encouraged to redistribute the sand of the Zandmotor along the coastline, thereby broadening the adjacent foredunes and beach. Specifically, the main objective of the project was defined as: 'Encouraging natural dune growth, primarily in width, in the coastal cell between the cities Rotterdam and The Hague. This 5 creates a larger sand buffer to cope with rising sea level as well as more space for nature and recreation and a larger freshwater lens under the dunes' (Fiselier, 2010;Van Slobbe et al., 2013).
However, while the Delfland coast partly maintains relatively wide and natural dune areas, the aerial photos in fig. 1 clearly show that in some places the dunes are not more than a narrow foredune ridge that is directly bordered by urban areas.
Because the region is densely populated, the coastline (including the Zandmotor) faces persistent pressure from anthropogenic 10 disturbances. Even though the Delfland coast is meant to serve a wide range of socio-economic functions related to recreation and leisure, activities arising from these function are often in direct conflict with the objectives related to coastal safety and natural values (e.g., Jackson and Nordstrom, 2011;Lithgow et al., 2013). Within this context, this paper examines to what extent the Zandmotor has contributed to creating accommodation space favorable for dune development, i.e. accommodation space that is sheltered from frequent storm impact and experiencing a steady accumulation of wind-blown sand. This favorable 15 accommodation space is identified, using publicly available remote-sensing data, by (1) comparing the presence of existing dunes against recent morphology of the Delfland coast and by (2) comparing recent coastal morphological changes against changes in dune cover by marram grass. Then, by taking into account the existing dunes and the conditions required for successful marram grass establishment, the identified favorable accommodation space is used to indicate the full potential for dune development along this urbanized and frequently nourished coastline. 20 FIGURE 1 ABOUT HERE.
2 Materials and methods 2.1 Regional setting The Zandmotor is located along the Delfland coast, an approximately 15 km long stretch of coastline that runs between Rotterdam and The Hague parallel to the dominant south-western wind direction. The Delfland coast has a long history of 25 coastal erosion; early 17th century maps make clear that the coastline, compared with today, experienced a significant (> 1.5 km) landward retreat (Van der Meulen et al., 2014). In the late 19th century the Delfland coast was fortified by groynes, but that only slowed down coastal erosion to a landward retreat of about 1 m/y on average. Therefore from the early 1980s onward, well before it became central policy, the Delfland coast has been frequently replenished with sand nourishments of varying volumes (Van Koningsveld et al., 2007). Still, in 2002 the Delfland coast was labeled a 'weak link' as it did not meet stricter coastal 30 safety standards that reflected expected increases in storm surge frequency and magnitude due to climate change (Keijsers et al., 2015b). Between 2009 and 2011, to guarantee for the Delfland coast to withstand hydraulic boundary conditions with Tucker, 1979). This ratio takes advantage of the contrasting reflection of photosynthetically active vegetation at visible and near-infrared wavelengths and is widely used for detection and classification of vegetated areas (e.g., Pettorelli et al., 2005;Nolet et al., 2017).

Linear spectral unmixing
The four selected bands were stacked into a new multispectral data cube and a linear spectral unmixing procedure was applied.

5
This was done to derive sub-pixel proportions of dune cover by marram grass, the most prominent and abundant dune-building species. Linear spectral unmixing is an approach to determine the relative abundance of user-specified ground cover components (endmembers) in multispectral (or hyperspectral) imagery based on its spectral characteristics (e.g., Smith et al., 1985;Settle and Drake, 1993;Theseira et al., 2003). It has successfully been applied before by Lucas et al. (2002) and Zhang and Baas (2012) in mapping the abundance of vegetation, including marram grass, in coastal dune environments. The reflectance 10 at each pixel of the image is assumed to be a linear combination of the reflectance of the endmembers present within the pixel: Where i = 1, ..., m are the number of spectral bands, R i is the reflectance of band i of each pixel, k = 1, ..., n are the number of endmembers, f k is the proportion of endmember k within each pixel, R ik the spectral reflectance of endmember k within each pixel on band i, and e i is the residual error term (Lu et al., 2003). Here, two endmembers were specified (see Fig. 2). The 15 first endmember was made up by a group of pixels (∼ 8) containing only beach sand, the second endmember by a similarly sized group of pixels fully covered by marram grass. The spectra of the two endmembers were obtained for each Sentinel-2 image separately, and maps containing sub-pixel proportions of beach sand and marram grass were derived using ENVI version 4.8 (Exelis Visual Information Solutions, Boulder, Colorado). The sub-pixel proportions of marram grass were subsequently interpreted as a percentage dune cover within each 10 meter pixel. Older established dunes (with NDVI > 0.6 in Fig. 2) were 20 excluded from the analysis as they are minimally exposed to marine forces and mostly covered with vegetation species other than marram grass. Further, all man-made structures on the beach related to coastal safety (e.g. groynes) and leisure and recreation were masked from the imagery before the linear spectral unmixing procedure was executed.
Changes in dune cover by marram grass along the Delfland coast were obtained by subtracting the percentages dune cover calculated for the 2016 Sentinel-2 image from the snapshot of 2017. Changes in dune cover between 2016 and 2017 were expressed for every 10-meter pixel but also as an alongshore change in cover area (m 2 /m/y). This was done for better interpretation of dune dynamics along the Delfland coast and was calculated by multiplying the surface area of each pixel (100 m 2 ) by its fractional cover change. The linear spectral unmixing procedure was validated against a high-resolution orthomosaic of a 5 strecth of foredune directly adjacent to the Zandmotor (see also fig. 3). The georeferenced orthomosaic (5 cm pixel size) was obtained by an Unmanned Aerial Vehicle (UAV) during a flight on September 1 2016, so 10 days before the acquisition date of the 2016 Sentinel-2 image. Using a k-means clustering algorithm (Hartigan and Wong, 1979), the individual 5 cm pixels of the orthomosaic were classified either as beach sand or marram grass. The accuracy of the algorithm was confirmed by visual inspection; for more details about acquisition and processing of the UAV-derived data the reader is referred to Nolet et al.

10
(2017). The orthomosaic was subsequently resampled to match the 10 m pixel size resolution of the Sentinel-2 imagery and dune cover depicted in the orthomosaic was calculated as the proportion of the (former) 5 cm pixels classified as marram grass contained within each newly aggregated 10 m pixel.  fig. 1) running along the crest of the newly created foredune ridge. In order to compare the morphology and morphological changes of the Delfland coast against the presence of dunes and changes in dune cover by marram grass, the 2 m resolution DTM's were resampled using bilinear interpolation to match the 10 m pixel size of the Sentinel-2 imagery. Figure 3 shows the results of validating the linear spectral unmixing procedure on the Sentinel-2 images. The dune cover calculated from the orthophoto and the two Sentinel-2 images are plotted against each other in fig. 3B. It is clear that deriving 5 sub-pixel proportions of dune cover using linear spectral unmixing result in an overestimation of dune cover by marram grass.

Results
Even though 54% of the variance for 2016 Sentinel-2 image can be explained by a positive linear regression model, most of the data points deviate from the 1:1 identity line because of higher dune cover values calculated by the Sentinel-2 image.
This trend, however, appears to be systematic to the linear spectral unmixing procedure since the data points from the 2017 Sentinel-2 image deviate even further from the identity line. This lower correlation (R 2 ≈ 0.35) is in line with expectation as 10 dune cover by marram grass was observed to have increased at this location between 2016 and 2017. So even though the linear spectral unmixing procedure overestimates the sub-pixel proportions of dune cover, the derived marram grass cover values for each Sentinel-2 image appear to be comparable relative to each other.  average the Delfland coastline has been accretive, at a rate of 19.3 m 3 /m/y, but it is clear that there has been a high alongshore variability in sand accretion and erosion rates. This can be attributed to the anticipated behavior of the Zandmotor: the accretive areas on its flanks gained approximately 2.8 10 5 m 3 /y of sand, while the erosive areas on its base lost about 1.7 10 5 m 3 /y of sand. Further, the foredunes experienced accretion of sand at relatively stable alongshore rates. In total, at an average rate of 11 m 3 /m/y, the foredunes along the Delfland coast gained approximately 1.6 10 5 m 3 /y of sand between 2013 and 2017.

30
The third map (4D) shows how the dune cover by Marram grass (expressed as percentage per 10 meter pixel) changed along the Delfland coast between the acquisition dates of the two Sentinel-2 images. Using changes in marram grass cover as a proxy for dune development potential, it appears that in most places the dunes along the coastline have been expanding over the course of a year. This observation, however, must be considered with some reservation, as observed changes in marram grass cover may also have been due to denser or taller growth of marram grass and not because of actual lateral expansion. Having said that, map 4D suggests that particularly the embryo dunes have been expanding, from 3 to 5 ha between 2016 and 2017.
As a result, in 2017 about 17% percent of the dunes along the Delfland coast could be considered embryo dunes, of which 5 most developed naturally along the coastline. The foredunes, in contrast, appear to have experienced more spatial variation in marram grass cover changes. Map 4D suggests that, along most of the Delfland coast, the foredunes have been expanding between 2016 and 2017. However, especially at the dune toe and just leeward of the dune crest, the foredunes appear to have declined somewhat in cover. This decline is most apparent north of the Zandmotor, which is clearly reflected in the second graph (4E) that shows the alongshore yearly change in dune cover (m 2 /m/y) between 2016 and 2017. This northerly foredune 10 decline will be examined in more detail in the discussion, but it can likely be attributed to anthropogenic disturbances (due to recreational activities as well as nature management practices) and to the fact that this stretch of coastline has not been nourished with sand when the Delfland coast was reinforced between 2009 and 2011. All in all, data from Sentinel-2 imagery suggests that the dunes along the Delfland coast have been expanding between 2016 and 2017 at an average rate of about 11.2 m 2 /m/y. For the foredunes this amounted to an increase of dune cover of 42 to 54 ha. Though, as stated before, this may 15 be an exaggerated number as increase in dune cover by marram grass has likely not been due to lateral growth alone.

FIGURE 4 ABOUT HERE.
Accommodation space is considered favorable for dune development when it is (1) sheltered from frequent storm impact and (2) experiencing a steady accumulation of wind-blown sand. The first (boundary) condition is identified by comparing the presence of dunes along the Delfland coast in 2017 to the height at which they were located. As fig. 5A demonstrates, in 20 2017 there were no dunes present below a height of 1.6 m +MSL. All embryo dunes were located on the beach between 1.6 -6 m +MSL, while the foredunes were located at heights between 6 -14 m +MSL. This suggests, at least for 2017, that dunes along the Delfland coast were sheltered from storm impact above a height of 1.6 m +MSL. Accommodation space, as a result, is considered favorable for dune development above this boundary height. The second condition is identified (or verified as the positive effect of sand burial on marram grass vigor is well documented) by comparing the changes in dune cover by marram assume that this accumulation of sand occurred predominantly by aeolian forcing. In addition, fig. 5B shows that almost all dunes increased in cover by marram grass between 2016 and 2017. Overall this increase in cover was most pronounced for the embryo dunes, as the foredunes showed limited increase and even some decrease in dune cover towards higher changes in dune height. The largest increase in dune cover between 2016 and 2017, however, coincided with the same change in dune height (∼ 0.1 m/yr) for both the embryo dunes and the foredunes.

FIGURE 5 ABOUT HERE.
The identified accommodation space favorable for dune development (i.e. located above 1.6 m +MSL in height and experiencing a steady accumulation of wind-blown sand) is shown in map 6A. Including the parts that were already covered by 5 marram grass, it is clear that large areas along the Delfland coast were favorable for dune development in 2017. Especially the sheltered and accreting southern and middle part of the Zandmotor stood out for its large favorable accommodation space for dunes to develop. This is reflected more clearly by graph 6C, which shows the favorable accommodation space along the Delfland coast (in m 2 /m) as well as the potential for new dune development. As map 6B shows, this potential is calculated by subtracting the dune cover already present in 2017 from the total favorable accommodation space. Graph 6C makes clear that

Discussion
This paper examined to what extent the Zandmotor has contributed to creating accommodation space favorable for dune de-20 velopment along the Delfland coast. The results indicate that the Zandmotor itself provides the most favorable accommodation space, for it has large areas located above 1.6 m +MSL that on average experience a continuous accretion by wind-blown sand.
As such, the results highlight that the Zandmotor supports an especially high potential for new embryo dunes to develop as most of its accommodation space is located on the beach. This section examines the merit of the identified conditions for when accommodation space is considered favorable for dune development, as well as the merit of using favorable accommodation 25 space to indicate dune development potential. The latter is examined in relation to the design and intended dynamical nature of the Zandmotor, the conditions required for successful establishment of marram grass and the persistent anthropogenic disturbances along the Delfland coast arising from recreation and nature management practices.

Conditions indicating favorable accommodation space for dune development
Accommodation space is considered favorable for dune development when it is sheltered from storm impact and experiences a steady accumulation of wind-blown sand. The latter condition is not disputed, as the reinforcing feedback between the growth response of marram grass and burial by wind-blown sand is well documented (Huiskes, 1979;Disraeli, 1984;Maun and Lapierre, 1984;Van der Putten et al., 1988;Hesp, 1991;Maun, 1998) and recognized to be fundamental to coastal dune 5 development in temperate regions around the world (e.g., Baas and Nield, 2010;Durán and Moore, 2013;Keijsers et al., 2016;Nolet et al., 2017). The positive feedback mechanism originates from a trait that all beach grasses of the genus Ammophila possess, namely potentially unlimited horizontal and vertical growth through its rhizomes (Gemmell et al., 1953;Ranwell, 1972). Whether marram grass grows horizontally or vertically subsequently depends on the amount of wind-blown sand, which makes its so particularly advantageous to dune-building. After establishment, by seed or rhizome dispersal, marram grass first 10 produces leafy shoots along newly developing horizontal rhizomes. When wind-blown sand is trapped by the leafy shoots, the immediate sand surface is raised and a small embryo dune is formed (Hesp, 1989). The leafy shoots are capable of growing up through a moderate thickness of sand by elongation of individual leaves. If, however, a leafy shoot is overwhelmed by sand deposition, one or more of its axillary buds develop into a vertical rhizome that will continue to grow until the surface is reached.
Adventitious roots are produced from the nodes of the vertical rhizome and the horizontal rhizomes gradually die, so that the 15 vertical rhizomes become independent of one another. This process may be repeated as long as aeolian supply is abundant and marram grass continues to trap sand. The capacity to trap sand, as noted before, is enhanced by the growth response of marram grass to sand trapping, which introduces the positive feedback mechanism driving coastal dune development (Gemmell et al., 1953;Ranwell, 1972). Using very high-resolution data, Nolet et al. (2017) showed that marram grass on foredunes along the Zandmotor appears to thrive best under a sand trapping rate of approximately 0.3 meter of sand per growing season and 20 that marram grass can withstand sand burial up to 1 meter of sand. However, while this demonstrates how positive plantsand feedback steers dune development, it must be noted that the physical size of a developing dune and predominant wind regime also controls its morphology (Davidson-Arnott et al., 2018). As dunes grow, for example, a limit is imposed on its height because the wind force required to transport sand upslope increases significantly (e.g., Arens et al., 1995;Arens, 1996;Keijsers et al., 2015a). Coastal foredunes therefore tend to expand in width rather than height, which emphasizes the importance of the 25 wide favorable accommodation space the Zandmotor provides for foredune development.
The condition that accommodation space is considered favorable when it is sheltered from storm impact warrants closer inspection, because the impact of a storm surge depends both on the magnitude of the storm as well as the geometry of the beach (Houser et al., 2008). Wind stress due to atmospheric pressure differences drive storm surge levels and offshore wave conditions, but the vertical dimension of the beach profile, in particular, exerts great control on shoreline parameters such as 30 wave setup, swash and runup (e.g., Stockdon et al., 2006;Sallenger Jr, 2000;Ruggiero et al., 2001). This is significant because the dissipation of kinetic energy of breaking waves is responsible for the highest rates of coastal erosion and dune decline (e.g., Vellinga, 1982;Short and Hesp, 1982). However, while empirical models can calculate wave runup levels and wave breaking energy from parameters such as offshore wave conditions and beach profile (see Stockdon et al. (2006) and Sallenger Jr (2000) for details), those relations only return approximations as often not all required model input is available or because of inherent model uncertainties. Having said that, the results suggest that dunes along the Delfland coast are sheltered from storm impact above a beach height of 1.6 m +MSL. This finding is examined in relation to offshore sea water levels measured by a buoy in close proximity to the Zandmotor mega-scale beach nourishment. Figure 7 shows the probability density curve (which is bimodal because of tidal dynamics) of those sea water levels (in m ± MSL), measured every 10 minutes from 2011 until 2017.

5
Included are the instances when sea water levels exceeded the apparent 1.6 m +MSL boundary height for dunes to be sheltered from storm impact. It is clear from fig. 7 that this did not occur frequently, only during about 0.4% of the measurements. Those measurements, however, were relatively clustered together, meaning that the boundary height was exceeded over (relatively) prolonged periods of time. Although, over the course of six years this happened for no more than 10 full days. On average the exceedance was about 20 cm up to a sea water level of 1.8 m +MSL, but on a few occasions sea water levels almost doubled 10 compared the boundary height to 3.10 m +MSL. This is excluding the wave runup onto the beach, which can be significant for natural beaches in the Netherlands. Dependent on whether the beach profile is dissipative or reflective, both Stockdon et al. (2006) and Poortinga et al. (2015) show that wave runup may reach to heights from 0.85 to 1.45 m above still water level (m), which is the level that would occur in the absence of waves. This implies that, since the construction of the Zandmotor in 2011, the Delfland coast may have experienced coastal erosion by storm surge levels reaching heights up to at least 4 m +MSL.
The observation that in 2017 quite a large number of embryo dunes were present on the beach at heights well below the maximum experienced storm surge levels, points to the capacity of established dunes to withstand and recover from hydrodynamic storm impact as well as to the pivotal role marine dispersal of rhizome fragments likely plays to dune establishment processes.
As remarked by various researchers (e.g., Suanez et al., 2012;Feagin et al., 2015;Houser et al., 2015;Van Puijenbroek et al., 20 2017a, b), the ability of embryo dunes to recover from storm impact largely depends to what the extent the above-and belowground structural integrity of marram grass remains intact after a storm event. This depends, in turn, on the severity of the storm impact on the dune, which can be caused by wave erosion (scarping and overwash) and swash inundation (Sallenger Jr, 2000; Hesp and Martínez, 2007). Wave erosion may completely remove all sand from an embryo dune (so it is no longer raised from the beach surface) and have an abrasive effect on the leaves of marram grass, causing either minor damage or compete 25 removal of all aboveground biomass. Most of the belowground root system of marram grass, however, has been observed to largely remain intact after wave scarping or overwash (Feagin et al., 2015). Potential damage of swash inundation to marram grass depends on the duration of the inundation period, but as Vergiev et al. (2013) demonstrate, marram grass displays no visible decomposition of stems, roots or rhizomes after being immersed with sea water for 20 days. This is well beyond the period a beach will be inundated after a storm event, which implies that inundation has a limited, if any, negative effect on throughout the year and often show a yearly net growth when aeolian supply was sufficient (Anthony et al., 2007;Suanez et al., 2012). This not only indicates that embryo dunes have the capacity to withstand storm impact and quickly recover to prestorm conditions, but also that the above-and belowground structure of marram grass often remains largely intact after a storm event.
Marine forcing, at the same time, has been shown to be an important agent in the dispersal of marram grass rhizome fragments and subsequent dune establishment via clonal growth (Konlechner and Hilton, 2009;Hilton and Konlechner, 2011). The distribution of (embryo) dunes on the southern part of the Zandmotor, as shown in close-up in fig. 8, suggests a correlation to . This fresh water availability, in combination with moderate burial dynamics, have been shown by Konlechner et al. (2013) to be beneficial to marram grass seed germination and subsequent dune establishment. The specific distribution of embryo dunes around the dune lake may therefore correlate best to seed dispersal by wind coming from the dominant south-western wind direction, either pushing the seeds over the lake towards the north-east corner of the lake or depositing it on the south-west lee side where the beach slopes downwards towards the lake. In effect, fig. 8 illustrates that, even 15 though the Zandmotor may provide wide favorable accommodation space and thus a high potential for dune development, the conditions required for successful (natural) dune establishment must also be considered. Having said that, Puijenbroek (2017) showed in a field transplant experiment that planted marram grass (consisting of a rhizome fragment with one shoot) thrived on most parts of the Zandmotor except when exposed to direct wave action. This suggests that conditions that limit marram grass growth and subsequent dune development (e.g. high salinity, drought, low nutrient status) are mostly absent on the Zandmotor 20 and likely along the entire Delfland coast.

Dune development potential in relation to anthropogenic impacts
The results highlight the overall importance of the Zandmotor for dune development along the Delfland coast. First, this is because its beach provides very wide favorable accommodation space that therefore supports a high potential for new embryo 25 dune development. And second, because of its sand feeding effects, the Zandmotor has likely contributed to creating more favorable accommodation space for dune development along the entire Delfland coast. The coastline directly north of the Zandmotor, for example, experienced a significant accumulation of sand between 2013 and 2017 even though it has not been nourished with sand in the years before. Although the amount of sand accumulation was less compared to the coastline that has been nourished between 2009 and 2011, the overall positive sand budget illustrates the intended dynamical nature of the 30 Zandmotor, where its sand is redistributed along the coastline causing a seaward broadening of the beach and dunes. In fact, graph 6C suggests that the unnourished northern part of the Delfland coastline supports a higher potential for dune development compared to the nourished southern coastline. In part this may be due to the fact that the Delfland coast is characterized by a net northward sediment transport regime (Van Rijn, 1997), which is reflected in the sand feeding budget of the Zandmotor. In the first 18 months after its completion, De Schipper et al. (2016) for example show that up to 40% more sand of Zandmotor was transported in a northward direction rather than southward towards Rotterdam harbor. At the same time, because the 2009 -2011 nourishment strategy consisted (for a large part) of foredune reconstruction that included plantings of marram grass, the created favorable accommodation space along the nourished coastline may not be accommodating to much new dune 5 development. As such, even though the coastline south of the Zandmotor has been reinforced with sand nourishments, it is quite possible for the unnourished northern Delfland coastline to experience more pronounced dune development in the years to come.
Interestingly, however, the positive effect of the Zandmotor on the northern Delfland coastline, in terms of sand accretion, is not reflected in the changes of cover by marram grass between 2016 and 2017. Even though it is shown that the coastline 10 north of the Zandmotor provides ample favorable accommodation space, it appears that the potential for dune development is currently not being realized. There are two main anthropogenic impacts that may hamper dune development along this urbanized coastline, namely persistent disturbances arising from recreation and leisure as well as a (increasingly prevalent) nature management practice that is aimed at remobilizing the dune landscape. Figure 9 gives an overview of total alongshore changes in dune cover by marram grass between 2016 and 2017 (in m 2 /m/y) and aims to relate it to anthropogenic activities that  fig. 9B shows, the seaward side of a beach entrance is commonly paved with concrete slabs and cuts relatively deep into the stoss slope of the foredune. This, effectively, mimics a through foredune blowout (e.g., Hesp, 2002), in which wind erosion is enhanced because of local wind speed acceleration and pronounced turbulent flow structures such as corkscrew vortices (Hesp and Martínez, 2007). Because the floor is paved, these wind-driven forces will in particular erode (i.e. widen) the slopes of the beach entrance and this 35 susceptibility to lateral erosion may have lead to the observed decline in marram grass cover. Second, as can also be seen in fig. 9B, there is often a hospitality establishment (e.g. a beach bar or restaurant) directly besides a beach entrance. And although their placement on the beach is often seasonal, their presence is numerous. In the summer of 2017, for example, only three of the twenty-three beach entrances along the Delfland coast did not have one or more hospitality establishments directly placed besides it. Perhaps not coincidentally, two of those three entrances gave access to the more isolated parts of the Zandmotor. The 5 presence of hospitality establishments puts additional pressure on the dunes as people may flock around the beach entrances and motorized vehicles are more common, for example to resupply the establishment. Even though walking or driving in the foredunes is prohibited along the Delfland coast, several studies (e.g., Andersen, 1995;Anders and Leatherman, 1987) show that vehicles and people on the beach may have a significant negative effect on dune development.
At the same time, as laid out in more detail by Jackson and Nordstrom (2011), the structure of the hospitality establishment Further, another important anthropogenic disturbance with a highly negative impact to dune development, is that the beach 20 directly north of the Zandmotor is mechanically raked during the summer to remove wrack line material and human litter.
Even though it is a common practice to accommodate beach recreation (Jackson and Nordstrom, 2011), this severely hampers embryo dunes from establishing themselves on the beach. Not only can the used machinery destroy any sprouting seedlings or rhizomes of marram grass, the removal of wrack deposits also deprives marram grass from potential hospitable locations to establish itself on the beach (Kelly, 2014). As a result, these anthropogenic disturbances combined have likely contributed to 25 reduced dune development compared to the rest of the nourished Delfland coastline.
Then, as fig. 9C and fig. 4E as show, the decrease of marram grass cover along the unnourished northern part of the Delfland coastline suggests that the foredunes have been in decline between 2016 and 2017. Which is unexpected considering the positive sand feeding effect of the Zandmotor on this stretch of coastline. Upon closer inspection, the main candidates for the observed foredune decline are a number of dune excavations aimed at rejuvenating the dune landscape. When the focus 30 of Dutch coastal policy widened, to also include preserving the spatial quality and natural values of the coastal zone, it was recognized that traditional flood safety measures had led to over-stabilized dune systems that were characterized by a markedly reduced biodiversity compared to younger and more dynamic dune systems (e.g., van Dorp et al., 1985;Provoost et al., 2011).
For that reason, in places where coastal safety could be guaranteed, remobilizing dune systems by removal of dune vegetation and topsoil has become a key management practice for maintaining a high biodiversity in the dune landscape. Reinitiating 35 aeolian dynamics is hereto essential, as deflation and deposition zones creates habitat diversity and renewed opportunities for specialized pioneer vegetation species (e.g., Arens et al., 2013). Nowadays, in order to maintain or even increase dune mobility, the rejuvenated dune systems are often connected to the beach and foredunes through the excavation of foredune notches. This has been shown to result in a sustained input of wind-blown calcareous beach sand and more diverse living conditions for pioneer vegetation, e.g due to higher levels of sand burial, wind speeds or salt spray (Riksen et al., 2016;Ruessink et al., 2017).

5
However, as fig. 9C shows, dune excavation practices appear to adversely affect foredune development. While there were no foredune notches explicitly excavated along the Delfland coast, fig. 9B shows that the paved beach entrances may similarly act as conduits for aeolian transport into the dune excavations. As a result it is quite possible that the narrow foredunes in front of the excavated dune are experiencing a net deflation of sand which negatively affects the growth of marram grass (e.g. by root exposure). Sand deposition then likely occurs deeper landward where no marram grass is presently growing to benefit from an  to 2017 to the average yearly change in dune height between 2013 -2017 demonstrates that dunes were almost exclusively 20 present in accreting areas. As such, even though its design may not be optimal for successful marram grass establishment, the results highlight the overall importance of the Zandmotor to dune development potential: -Compared to the rest of the Delfland coast, the supratidal beach of the Zandmotor provides very wide favorable accommodation space and therefore supports a high potential for new embryo dunes to develop.
-Because of its sand feeding effects, the Zandmotor will likely contribute to creating more favorable accommodation 25 space for dune development along the entire Delfland coast.
However, because of persistent anthropogenic disturbances arising from recreation and nature management practices, dune development along this urbanized coastline may not reach its full potential. This should not be too alarming, though, as the Zandmotor mega-scale beach nourishment is set to ensure the safety of the Delfland coast for years to come.

Competing interests. The authors declare no competing interests
Acknowledgements. The research was carried out within the program Nature-driven nourishment of coastal systems (NatureCoast) and