General comments

Overall this study is well designed and thought through. The topic of dune initiation and plant establishment is important for coastal research and management. Including both naturally occurring sites as well as an establishment experiment was a great idea to not only compare results, but for comparing managed vs. unmanaged dune systems. There are significant data needs in this community, and this research helps fill the gaps. These species specific data are novel findings nd relevant to the overarching body of knowledge.

Reply: We thank reviewer 1 for this positive and thoughtful feedback on our study.

Specific comments

One thing that’s missing from the introduction and discussion are comments on sea-level rise. This could be including the projected annual SLR for the sites in the methods. Which could be added to table 1. However, throughout the paper when discussing the other environmental factors affecting dune establishment, sea-level rise should be included. Section 5 of the discussion would also be a great place to add this information. 

Reply:  Indeed, including SLR is an important consideration, as it may lead to erosion of beaches and can reduce the accommodation space for dunes. To address this, we will add the impact of sea level rise to the introduction and additional literature to the discussion in section 5. We decided against adding the information in Table 1 because in the Netherlands, currently sea level rise is offset by intensive nourishment activities (Keijsers et al. 2015)  and therefore does not pose a direct problem for establishment and dune development. Rather, dunes are expanding seaward (van IJzendoorn et al. 2021).

The “window of opportunity” statements (including in the title) are a little misleading given the project. There are no phenology or timing of emergence data included, which given the title I would expect. The findings narrow down the drivers and limits of dune initiation for these species. I’d recommend changing the wording throughout to remove “window of opportunity,” or define it clearly. There is a definition on line 42, however the findings are more about the ongoing environmental factors and not disturbance. The definition would need some clarity. 

Reply: We agree with this assessment. Indeed, the aim was not to narrow down timing or phenology of the emergence data but rather to quantify environmental conditions that explain long term establishment success and subsequent dune initiation. While the window of opportunity term could potentially be defined in a broader sense, we will remove it from the manuscript to avoid further confusion. We propose to change the title to “Narrowing down dune establishment drivers on the beach” and replace the term “window of opportunity” by “drivers” throughout the manuscript.

It’s mentioned in the discussion, but along with Table 2 in the methods, it would be good to include a seed size comparison of the 2 species. Table 2 shows the species differences, but given the objectives of the study this is an important point to include earlier in the text. 

Reply:  We tried to mimic the natural dispersal of both species. While we introduced both species as spikelets it is important to note that Ammophila arenaria spikelets typically contain a single floret (Huiskes 1979), while Elytrigia juncea contains several florets within a dispersed spikelet. This makes an important difference for the size and weight. In light of this comment and to further clarify we will highlight this aspect better in the discussion while also adding measurements on the seed and spikelet size to table 2 and a photo of the introduced plant material to the supplements.

In Figure 2A it is difficult to distinguish between the 3 levels of “establishment success”, I’d recommend using 3 different shapes instead of different sized circles. 

Reply: The establishment success in Figure 2A is displayed as a continuous variable rather than a variable with different levels. The size of the circle is relative to the establishment success. To clarify we are planning to adapt the figure legend and circle sizes by changing the value range for clarity.

Technical corrections 

Line 38: Used common name when throughout the rest of the paper the scientific name is used. Both are listed in Table 2, but not in the text.

Reply:  We will change to the scientific name with the common name in brackets.

Line 203: In section 3.3 of results there is a switch from “monitoring moments” to “monitoring campaigns” - consistency would be helpful here. 

Reply:  We will change “campaign” to “moment”.

Line 246: I think you mean ‘wrack’ not ‘wreck’

Reply: We will correct “wreck” to “wrack”.   

General comment

The authors have chosen to study an important topic. We have a limited understanding of the controls on seedling emergence and subsequent dune initiation. While the topic is important, I do not believe that the methods employed as described are at a collection interval that is appropriate for the hypotheses. The abiotic conditions being measured are collected and related to seedling emergence, but the conditions occurring within the time between the collections/emergence which impact if not control emergence triggering are not considered/measured. For example, dormancy is broken in many plants by a combination of suitable conditions being met across a critical number of consecutive days to trigger emergence – the same is to be expected here, but not accounted for in how the work is designed and then discussed. There is a major assumption that the difference between the conditions between collections (which appears to be month to month+) drive the emergence. This is a huge assumption that would need to be justified by the literature or acknowledged in an honest way with the discussion.  With this in mind, the meaning of and implications of the results are hard to interpret. This is a biological or ecobiological topic, but the biological does not appear to be the focus in how the methods were designed (timescale) or discussed (emphasis of the context of the work) making it unclear if the results mean anything and if they do how they add to our previous understanding of this system. The methods must be improved to understand the reliability and meaning of this work.

Reply:  We thank Reviewer 2 for their thoughtful and constructive feedback, as well as for recognizing the relevance and importance of our study. We appreciate the critical concerns raised regarding the methodology and the relationship between the environmental conditions measured and the ecological responses observed. We acknowledge that a clearer definition of the ecological processes we were measuring would improve the manuscript, particularly regarding the interpretation of seedling emergence and subsequent dune initiation.

The aim of our experiment was to examine establishment from a realized niche perspective over a time scale relevant to dune initiation. As the reviewer correctly points out, there are multiple environmental and biological filters that influence plant success at different life stages, starting from germination (which requires dormancy to be broken), to survival and growth. All of these life stages are expected to be influenced by the measured environmental conditions, which is supported by the literature cited in our introduction. However, it is important to clarify that our study does not explicitly separate these stages, and the shoot numbers observed in the field represent the net result of these processes. We are planning to highlight this aspect better throughout the manuscript by adjusting the definition of the ecological process.

We also acknowledge the reviewer’s concern regarding the temporal resolution of our data collection. While temporal conditions can influence processes like dormancy and germination, our study was primarily focused on the role of spatial differences in environmental factors, summarizing results over 750 plot replicates. While moisture, salinity, and bed level changes vary over time, they are also strongly influenced by spatial differences, as discussed in the introduction (line 55). To address the limited temporal resolution of our study, we averaged all variables over time to capture broader trends (line 145, Methods). We will revise the discussion to stress our focus on spatial patterns and acknowledge limitations of low temporal resolutions.

Intro

• Line 25 - is misleading and should be reworded or a citation should support the claim that nourishment “is used for…stimulating new dune development and improving the climate-resilience of dunes.” Dunes are often not included in nourishment designs at all and so while net positives in sediment supply to the dunes may occur and can be a benefit, this occurring is an unintentional benefit.

Reply:  We will adapt the text highlighting that nourishments are not carried out to create dunes but rather dunes can result from nourishments.

Methods

• Methods are missing the ‘study site’ equivalent for relevant information on the biology of the two study species. This would improve putting the work into context. 

Reply: We appreciate the suggestion and we will include additional information on the ecological differences between the two study species in the methods section. Specifically, we will highlight that Elytrigia juncea is a pioneer dune-building grass commonly found in the embryo dunes of Northwestern Europe, whereas Ammophila arenaria typically occurs at later successional stages within foredunes. Its clumped growth form contributes to the formation of taller, steeper dunes compared to Elytrigia.

• Study site is missing grain size information which is relevant for interpreting dune initiation results around critical wind speeds and angle of repose for expected morphologies and sediment supply

Reply: We will include grain size information for both study sites.

• I have a lot of questions about the experimental design relative to the driver being tested and understanding the relevance of the results. How are the authors accounting for differences in abiotic conditions across the four areas, 2 sites, and then within them?
• In S2 the authors define the environmental conditions as ‘similar,’ but this needs to be defined in the actual methods and more specifically as it is critical to understanding how the experimental design may impact the results.
• The authors are citing Houser and so must be aware that there are differences in abiotic conditions relevant to plants that can occur on the order of meters and must be accounted for in experimental design (by limiting plots to close quarters, using a paired design, explicitly monitoring the driving conditions daily, etc).
• As written, I understand that the plots are block designed and 1 meter apart from the next closest, but across what span in the similar areas for which there appear to be 10 per table 3 – the area might be ‘similar’ but plots 50 m apart in the same area on opposite ends are likely different so that data should be presented that shows the abiotic conditions are comparable as measured in the field beyond TWI and DTM (presented in supplementary). I assume that the authors have accounted for all of this, but this needs to be made clear and so methods improved for clarity and reproducibility. How can the authors convince a reader that differences in abiotic conditions are equivalent and so what they are presenting as results are as they describe relative to their hypotheses as opposed to just a documentation of inherent beach/dune heterogeneity?

Reply: In our study, we addressed potential confounding effects both within and across study areas through a systematic experimental design and the inclusion of random effects in the statistical analysis.

For the experimental design, we adopted a stratified approach: We first delineated areas (i.e. strata) with similar expected environmental conditions, based on relevant environmental proxies (TWI, change in elevation and height above mean sea level), and then allocated an 15 blocks per area and study site. Within each strata (or similar area), we further distributed block locations across both coordinate and covariate spaces using the method of Grafström and Tillé (2013). This spatially balanced sampling minimizes potential impacts of spatial autocorrelation, improving our estimates. Our aim was not to ensure that environmental conditions were identical but to ensure that a broad range of environmental conditions were represented across all study sites, while also covering a typical range within each individual site.

In the statistical model, which forms the basis for the results reported, we accounted for potential differences across and within sites by including random effects. These random effects were applied at both the study area level (site-specific) and at the block level (local random effects). This allows us to statistically control for the random variation introduced by unmeasured variability arising from spatial differences at these levels and isolate the effects of the abiotic drivers under investigation. In doing so, we ensure that our results capture broader patterns rather than being overly influenced by site-specific conditions.

We will further clarify and elaborate on this methodology in the methods section of the paper, making sure that all of the points mentioned above are included. We will address these concerns by:

1. Providing more detailed explanations on the stratified design (e.g. how areas were delineated)
2. Providing additional information on the doubly spaced method for spreading the blocks in coordinate and covariate space (currently included in the supplementary material)
3. Extending the explanation of the random effects and how they could account for within and across site specific differences in abiotic conditions for the statistical models (currently only briefly stated on line 160)

How is the position of the plants in turn impacting shielding and so sediment supply and elevation change measured being accounted for in the block design?

Reply: We recognize that the positioning of plants within each plot could impact shielding effects. Especially in the case of plots located downwind we expect a possible influence on the sediment supply and the elevation changes upwind.

In our block design we aimed for controlling spatial variability, and we allocated each treatment only once per block. While we considered spacing the plots at a greater distance, practical constraints—such as avoiding pre-existing vegetation and pathways - meant that a minimum spacing of 1 meter was the only feasible option.

Despite the mostly sparse vegetation within the plots (approx. 12 cm in height, with an average of 23 shoots per 50 x 50 cm plot), vegetation still may influence wind velocity and sediment transport. Wolfe and Nickling (1993, 1996) discuss that even sparse vegetation can lead to wake interference flow. This may particularly be relevant in areas with more pronounced plant growth. Although vegetation development varied across the plots, with only some showing significant shoot growth across all plots, it is possible that these wake effects altered sediment deposition patterns and influenced elevation changes.

Moreover, Hesp et al. (2019) showed that, even with low vegetation cover (~10–12%, approx. equivalent to 90 shoots/m²), bed level changes can occur at a distance of more than 2 meters from the vegetation downwind. Therefore, we anticipated that within a single block, plots positioned downwind from those with higher vegetation development to possibly experience increased sediment deposition and elevation changes.

To address those plot-to-plot interactions, we incorporated the following strategies:

• Plot Orientation: Plots were oriented towards the North to standardize wind exposure.
• Randomization of Treatments: We randomized treatments within each block, ensuring that no treatment was systematically positioned downwind or upwind. This meant that any effects of vegetation shielding or sediment redistribution were randomized across treatments.
• Statistical Controls: To account for the potential influence of nearby plots within the same block, we used random effects per block in our statistical analysis, thereby managing within-block correlations.

While we cannot completely rule out interactions between plots—especially regarding sediment redistribution - we did not observed dune tail that extended into other plots. We will incorporate these considerations into the method and discussion section of the main manuscript.

• The height differences are being ascribed to the plants but increased deposition or sediment supply can trigger plant response such as emergence resulting from in increased retention. It is just as likely that differences in burial are driving emergence, survival, and density, as it is that changes in the emergence, survival, and density in turn impact the burial – this does not need to be tested necessarily, but I do not see it being addressed/discussed despite its importance.

Reply: We thank reviewer 3 for raising this is an important point to discuss. Indeed we cannot separate the effect between the two directly. It is likely that both factors mentioned here play a role. We will expand on this in the discussion, highlighting that burial may drive emergence while also the number of emerged shoots can influence sediment trapping.

• Line 122 – “We then corrected this number for spontaneous shoot emergence in the control plots.” Great, this is important to control for. How was this done so that others can repeat the methods?

Reply: We subtracted the number of shoots established in control plots from the number of shoots established in the treated plots, setting shoot numbers to 0 where the control exceeded the treatment. We will add a sentence to clarify this in line 122 of the methods.

• In line 125 establishment is defined as “We expressed plant establishment success as the corrected number of shoots present in a plot at the last monitoring round relative to the amount of introduced plant material in March 2022.” Please rework this definition to include info accounting for emergence from seed and rhizome being studied.

Reply: We will correct this and other definitions to highlight that the establishment success measured includes, survival, mortality, germination and growth, without explicitly separating the phases.  

• Line 128 - bed height was measured looking at two plot corners. I am assuming that this is 2 of 4 examined, but what two relative to the predominant wind direction? If you collected upwind corners from some and downwind from others then they are not necessarily comparable (apples to apples). Were the same corners measured each time? If you collected upwind corners one collection and then downwind another and took the difference then you would almost assuredly see a net increase in height because of shielding and nebkha formation. I assume all this was accounted for, but it needs to be stated for reproducibility and clarity.

Reply: The main wind direction in the Netherlands comes from the SW, therefore we opted for measuring the elevation at the SE & NW corners. We always measured the same corners during repeated data collections. We will add this information to the measurements section of the materials and methods.  

• It is not clear how maximum salinity and max moisture measurements were used in analyses and if this is max per plot, area or site. In general, it is hard to understand what data was collected and from where each of the four collections,

Reply: During each monitoring campaign measurements were taken for all five plots per block. This includes shoot counts, soil moisture, salinity, and the height of the NW & SE corners. Measurements were not taken per block but rather per plot. Hence, at each measurement campaign we re-visited all plots and measured in 750 plots. The only exception for this is the last measurement campaign where only blocks were re-visited that contained shoots at any point of the measurements (see line 120, i.e., 635 plots). We will name the plot number and measurements more explicitly in the methods section to clarify. When it comes to the maximum salinity, minimum soil moisture and maximum bed level change we used the values measured per plot and monitoring period. Reviewer 3 pointed out that this appears to be a minor point as these conditions should be related with the average conditions measured. We will therefore remove this analysis.

• I do not believe that the authors collected salinty, elevation and moisture data from about 700 plots each collection, but that is what i'm assuming because of how the methods are presented. Was salinity, height, moisture measured from each block, plot, area? A paragraph at the start of the methods making it clear what was collected where and when across all data would be very helpful to improve clarity.

Reply: The assumptions of the reviewer are correct: it really involved gathering data from 750 plots per collection round (only exception see reply directly above). The experiment was created to capture spatial differences and therefore carried out over a large spatial scale and did indeed involve substantial data collection efforts. We will make sure to mention more explicitly the scale of the study in the methods.

• It is not clear how often data were being collected from the field. I am assuming the data mentioned in section 2.3 was collected in 4 collections, in 2022 once in May/June, August and October.

Reply: It was indeed collected during 4 collections, once during May/June, once in August, once in October and the final collection was taken during January/February of 2023. We will adjust the description to be more explicit.

• In general, without this information it is hard to understand if the methods are appropriate. If the authors only collected this data 4 times then I have concerns about if the experimental design is appropriate for the hypotheses; the authors are examining salinity, bed-level change, and soil moisture, all factors that can vary daily with abiotic conditions and especially related to wind events, but it seems they were only collecting data at inconsistent month or month+ long intervals and not associated with events so that the drivers being tested are not actually being linked to the changes in vegetation in a real way that does not rely on an assumption that correlation means causation. This would have been a great application for the use of iLoggers.

Reply: Our study focused primarily on capturing spatial differences in environmental factors across sites rather than the short-term temporal variability. The primary goal was to assess how these spatial gradients influence vegetation over longer periods. We acknowledge that processes such as wind-driven bed-level changes or moisture fluctuations can impact results on shorter timescales, but we believe that the spatial heterogeneity we captured (with 750 plot replicates across diverse environmental conditions) offers a robust foundation for identifying patterns. We focused on capturing spatial rather than temporal variation given the strong spatial zonation of vegetation patterns on the beach.

To address potential limitations from lower temporal resolution, we averaged key variables (such as moisture and salinity) over time, which helps capturing broader trends. In the revised discussion, we will better emphasize that this study was designed to reflect spatial variability and that further research may be needed to address short-term temporal dynamics linked to specific weather events.

• Line 135 - this section/paragraph on seedling mapping is not clear in a way that can be replicated. As worded, it appears that the plant densities were estimated as opposed to quantified   

Reply: We will rephrase the section better separating the interpolation (used for visualization) and the calculation (estimate of seedling densities).    

Results

• I do not see that the results are compared across sites. It seems like all the data are lumped together across 4 areas from 2 sites, but I am not convinced this is appropriate based on the info provided in the methods to justify it. Site differences can confound these results so that the significance being ascribed many just be inter-are or inter-site as opposed to a function of a real overarching pattern observed consistently relevant to all three hypotheses.

Reply: We will extend the information on the random effects as part of the statistical model descripition.

Line 199- What are “extreme conditions” I do not see that these are defined in the methods or supplementary.

Reply: See elaboration in earlier reply. This is not essential to the manuscript and we opt to remove it.

Pictures of the plots as a figure would be helpful to see and make it clear how different planted areas vary from the controls where seeds and rhizomes were not planted.

Reply: We will add some pictures to the supplementary material showing differences between the different plots.  

Discussion

• The results are not surprising, which is not an issue, it’s just that they are seem largely restated in the discussion as opposed to in context to explain them with the existing state of the science/literature, section 4.1 is an exception to this statement.

Reply: While there is considerable amount of literature on the growth response of adult dune building grasses, the amount of literature on the establishment of juvenile dune-building grasses is surprisingly limited. We expect the response of adult grasses to differ from the response of juvenile dune-building grasses which makes the comparison between literature values not straightforward. Having said that, we will add some additional references to the discussion to give some reference values.

• Findings re-hypothesis #1 this i emerge close to their parents. There is a whole body of literature around this not addressed or discussed, both in dunes and not in dunes

Reply: We thank reviewer 2 for this insightful feedback. We agree that seedling emergence close to adult plants has been well established within ecological studies both at coastal dunes as well as in other ecosystems. This is precisely one of the reasons why we created this experiment: It is generally difficult to separate dispersal limitations from environmentally imposed establishment limitations. By introducing plant material across the study areas, we effectively removed potential dispersal limitations, allowing us to focus on the role of environmental constraints. By comparing the patterns between spontaneously established seedlings and seedling emergence from introduced seeds we can also quantify how big the dispersal limitations are relative to environmental limitations, something that has not been done before.

Specifically in dunes, Konlechner is cited, but then this work is not put in context with it or other similar works other than saying vaguely that the results are “in line with other studies.” Reijers et al. 2021 is also cited in the intro, and this work is very relevant here, but never discussed in the discussion

Reply: The study by Reijers et al. 2021 focuses on how the clonal expansion strategy of dune-building grasses can be influenced by burial from sediment. Clonal expansion is one of the main mechanisms for colonization of new areas on the beach. In our study we found that  establishment from seeds does not seem to be a main factor for marram grass, but for sand couch. This may mean that marram grass dominantly colonizes by clonal strategies. We will highlight this better in our discussion section.  Furthermore, we will expand on the study by Konlechner et al. 2013 giving reference values for rhizomes buried in glasshouse experiments (under static conditions) and the effects of storms on naturally dispersed rhizome fragments of Ammophila arenaria.

We will revise the discussion to more clearly highlight the novelty of our study. Specifically, we will emphasize the following key points:

1. Separation of arrival limitations and environmental constraints: Our study uniquely isolates these factors by introducing plant material on a large scale, allowing us to assess establishment success without the confounding influence of seed dispersal limitations while contrasting this with natural spatial emergence patterns.
2. Species-specific establishment differences: We identified significant differences in establishment success between the two species, which may be attributed to either arrival dynamics or varying resilience to environmental conditions. These differences provide insight into the species’ niche differentiation within dune systems.
3. Plant density dependence in dune initiation: We found that dune initiation occurs only when environmental conditions favoring establishment promote sufficient shoot development.

By distinguishing arrival from environmental constraints and exploring these species-specific patterns, we believe our findings contribute novel insights to the field, particularly in terms of understanding the factors limiting dune initiation.

• Findings re-hypothesis #2: I would suggest that the authors remove this hypothesis and element of work from the paper as the methods and data presented are not adequate to test this in a meaningful way as presented.

Reply: We disagree with this assessment. While daily or event-driven fluctuations could indeed provide additional insight into certain abiotic drivers, we believe the spatial heterogeneity captured in our study provides a robust method for identifying meaningful long-term patterns in vegetation response. To address any potential limitations, we averaged key variables (e.g., moisture, salinity) across time to reflect overarching trends rather than short-term variability. In the revised discussion, we will clarify the scope of our study—highlighting its focus on spatial patterns and acknowledging that future research with higher temporal resolution or event-based monitoring could complement our findings and provide a more granular understanding of temporal dynamics.

• Findings re-hypothesis #3: this is very interesting and holds implications for management and modeling of these environments. However, the few recent works on dune initiation (examples of main ones include Hesp et al. 2021, Hesp et al. 2019, Charbonneau et al. 2021, Costas et al. 2024) are lacking from the paper.

Reply: We thank the reviewer for the suggestions. Based on the comment we will expand the discussion on dune initiation while also taking a closer look at the recent study by Costas et al. 2024, cited on line 245. Context with regard to the following studies will be added:  

• Hesp, P. A., Hernández-Calvento, L., Hernández-Cordero, A. I., Gallego-Fernández, J. B., Romero, L. G., da Silva, G. M., & Ruz, M. H. (2021). Nebkha development and sediment supply. Science of the Total Environment, 773, 144815.
• Charbonneau, B. R., S. M. Dohner, J. P. Wnek, D. Barber, P. Zarnetske, and B. B. Casper. 2021. Vegetation effects on coastal foredune initiation: Wind tunnel experiments and field validation for three dune-building plants. Geomorphology 378:107594.
• Hesp, P. A., Y. Dong, H. Cheng, and J. L. Booth. 2019. Wind flow and sedimentation in artificial vegetation: Field and wind tunnel experiments. Geomorphology 337:165–182.

• Here a bit oor egg first style discussion would be interesting and relevant. What are next steps? How could the experiment be improved? What don’t we know still and as a result of this study what do we now know around this? 

Reply: As mentioned in previous comments we will expand on the discussion on the “chicken and egg problem”. We will further also address some additional study limitations (e.g., as outlined in previous replies) and possible next steps.

• The differences in seedling and rhizome success are the most relevant and novel findings. I would suggest leaning into this more in how the work is presented/framed relative to Hyp 1 and 3.

Reply. We thank the reviewer for this suggestion and will highlight the difference in establishment from seeds and rhizomes in the discussion and introduction.

Reviewer 3:

Homberger and colleagues collected an impressive dataset to understand environmental drivers of the establishment of two important ecological engineers from dunes: Elytrigia juncea and Ammophila arenaria. They combine censuses of plant emergence with supplementation experiments to understand how sand burial, erosion, but also how (average) conditions of humidity, salinity affect growth of the plants, and their subsequent impact on dune development. The strength of the work is clearly in the high spatial resolution at four beach sections in the Netherlands. I found the paper overall well written from a textual perspective, but especially in the methods section unclear to fully grasp what has been measured, let alone how the data were analysed.  The work is clearly hypothesis driven (last paragraph of the introduction) but i am unsure (because of the very limited description of methods and results) whether the dispersal limitation can actually be tested.  This is important as some strong conclusions are drawn from this (lines 295 and further). In conclusion, this seems to be a very good study, but it needs revision to reach its full potential with respect to impact and reproducability

Reply: We thank Reviewer 3 for this thoughtful feedback and the critical considerations. We especially appreciate the acknowledgement of the efforts regarding the dataset collection. In line with Reviewer 2, we understand that the method description could receive more attention to detail and we are planning to address this.

Detailed comments:

Line 64: windows of opportunity typically refer to temporal windows where sudden environmental conditions are favorable and promote establishment. Here, it is more used from a realized niche perspective as i could not detect any proper analyses of the temporal effects. Only in S7 some data are provided, but they are clearly not central to the work as only rudimentary touched upon.

Reply: In line with our reply to reviewer 1, while the window of opportunity term could potentially be defined in a broader sense, we will remove it from the manuscript to avoid further confusion. We propose to change the title to “Narrowing down dune establishment drivers on the beach” and replace the term “window of opportunity” by “drivers” throughout the manuscript.

Line 120. Only plots where shoots were present the year before were visited to record dune development. How can you separate whether the shoots accumulated sand, or whether shoots were already shooting on elevated locations, so on embryonic dunes.

Reply: Dune initiation was measured relative to the surroundings of the plot. So it is indeed possible that some of the locations were already elevated. Moreover, it is possible that shoot emergence happened as a consequence of added sediments, or burial under optimum conditions. On the other hand, it may also be that more shoots emerged and therefore more sediment was trapped. Our experiment does not allow for a direct separation between the two aspects. We think it is likely that both play a role. To address this, we will add to the discussion on the feedback between burial and emergence.

Line 135: it is not clear how you can separate dispersal limitation (false absences) from establishment limitations (true absences from environmental constraints).

Reply: In the case of spontaneous establishment, we acknowledge that it is difficult to separate dispersal limitations from establishment limitations. This limitation is one of the reasons why we designed the experiment. By introducing plant material across the study area, we effectively removed potential dispersal limitations, allowing us to focus on the role of environmental constraints.

While we cannot completely rule out that observed seedling patterns may still be influenced by environmental factors, our approach demonstrates that the space for establishment could potentially be broader if dispersal limitations were absent. This is illustrated in Figure 2.

Line 138: this does not make so much sense to me. How can you infer changes in densities (so within plots to my opinion) from changes in frequencies of plots with relative to without existing mature vegetation.

Reply: We thank reviewer 3 for pointing this out. Density is a poor choice of wording here. We will correct seedling “density” to seedling “occurrence”.

Line 140: unclear how you interpolate per block, and how to compare with the natural settlement. What spatial analyses were done? Some spatial correlations? This is interesting but not developed. What i understand is that you compare the spatial pattern in blocks with/without supplementation? Or comparing spatial patterns of natural emergence and experimental supplementations in the same block (but then how can you separate natural from assisted settling)

Reply: We thank the reviewer for their comment. For block interpolation, we converted the recorded shoot numbers of E. j. and A. a. to presence/absence data per block and used a simple interpolation method primarily for visualization, as shown in Figure 2. We didn’t develop this further because a formal spatial analysis would be complex: In the mapping approach, we used a higher plot density and larger plots compared to the introduction experiment, which makes direct comparisons challenging. Moreover, we did not quantify the seed or plant material bank for natural establishment, whereas in the establishment experiment, we precisely controlled the amount of introduced plant material. While this difference allowed us to examine environmental constraints by removing dispersal limitations in the experimental blocks, we cannot exclude this factor from the spontaneous establishment. We will add this to our discussion regarding study limitations.

Line 142: so you also are able to test seedling mortality? I overall don’t understand how the first approach (absence/presence after one growing season, and linked to max or min changes) is different than its correlation to average environmental conditions. I assume these are the same abiotic factors? Is plant establishment success then different that presence/absence after one season (not to me), or is the difference in the average versus min/max conditions? Also in stats, you mention absence/presence but never disappearance or conversely survival.

Reply: Indeed the difference is minor and we expect a correlation between average and maximum measured conditions. This was a minor analysis point in the main manuscript and we opt to remove it. Plant establishment success measured includes, survival, mortality, germination and growth, without explicitly separating the phases. We are planning to better highlight this in the main manuscript.   

Line 153: the k-clustering is used to identify plant establishment and dune development. I read further that it is used to identify or locate blocks?

Reply: k-means clustering was used in the experimental design to identify areas with similar env. conditions (or strata). As covariates we used TWI, elevation, and change in elevation. The blocks were allocated using doubly spaced sampling algorithm, spreading locations in coordinate and covariate space (same covariates). Both approaches were used seeking to capture a wide range in environmental conditions. A lot of this description was moved to the supplementary material. To further clarify we will move the design description to the main manuscript.

We also used the k-means clustering algorithm to identify areas suited for dune initiation and plant establishment. This was after the experiment had been conducted and the response and explanatory variables were measured. Variables we used included plant establishment success, maximum dune height, height above sea level, slope, and distance to the coastline at average high tide.

Line 156: which corners? The ones with strongest difference in height i presume?

Reply: The main wind direction in the Netherlands comes from the SW, therefore we opted for measuring the elevation at the SE & NW corners. We always measured the same corners during repeated data collections. We will add this information to the measurements section of the materials and methods. 

Line 158-168: the statistical analyses are too rudimentary explained in the main body. It seems that you control for spatial correlation between blocks and study area to remove spatial variation at larger spatial scales. I am still unclear how large blocks are so whether you have also within-block spatial variation that needs to be accounted for (for instance differences in shelter within blocks may promote dependency of responses in the plots). I would also suggest to document the error distributions for each model separately. From the supplements, i could first not understand which distributions were modelled, why some offsets for plant biomass were used. So this had to be better linked to different models (i.e. shoot establishment success is a proportion of the total added). In the model formulations, also explain what is s/ti is. Also, was model selection used or only full model considered.

I have the impression that analyses are done correctly but i remain unsure when linking the models to the different analyses, and how they are described.

Reply: We thank the reviewer for this suggestion. We accounted for potential differences across and within sites by including random effects at the study area and block levels, controlling for unmeasured spatial variability and isolating the effects of the abiotic drivers under investigation. Offsets were used to account for differences in plant material introduced, to model the shoot number relative to the introduced plant material.  We will further elaborate on the choices we made regarding the statistics in the main manuscript.

Line 175-178: where is this tested, why are these arrival limitations and not niche constraints?

Reply: In the case of spontaneous establishment, we acknowledge that it is difficult to separate dispersal limitations from establishment limitations. While we cannot completely rule out that observed seedling patterns may still be influenced by environmental factors, our approach demonstrates that the space for establishment could potentially be broader if dispersal limitations were absent. This is illustrated in Figure 2.

Table 4: why full models and salinity models? I somewhere missed why two different models were used here

Reply: This was explained in the supplementary materials. The effect of salinity was tested on smaller dataset, since the WET‐2 sensor which was available to us at the time is only able to record salinity once a soil moisture content of 15 % has been exceeded. We will move this explanation to the main manuscript.

Line 225: i need to be convinced of this. Is absence not a question of establishment limitation, or due to the fact that seeds are blown away, and accumulate close to sheltering adults plants?

Reply: Likely both play a role. We will extent the discussion regarding this aspect.

Line 253: this 2cm corresponds nicely with the depth the plants were buried. So this might be an artefact of the experimental treatment? I would assume, or learn from this that all seeds need some burial..

Reply: We thank the reviewer for pointing this out. The initial burial may have indeed played a role and we will add this aspect to the discussion. Nevertheless, both figure 3a) and b) point towards a tolerance against erosion that goes beyond the initial burial depth of 2 cm.

Line 330-333. Restoration projects typically start from shoots, so is this relevant?

Reply: Our findings should be indeed relevant for long-term restoration projects. Based on our results we would expect that Elytrigia juncea may have some advantages to establish as it can establish from seeds and rhizome fragments. Ammophila arenaria appears to be more dependent on rhizome fragments and clonal expansion. We will expand upon this aspect in the discussion.