the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 26 Jun 2025
Alongshore Varying Dune Retreat at a Barrier Island
Abstract. Barrier islands often exhibit spatially uneven morphological change due to alongshore variations in wave exposure, sediment supply, and dune morphology, which are influenced by both long-term evolution and short-term storm impacts. We investigate the alongshore variability of shoreline and dune evolution on Culatra Island (Ria Formosa, S. Portugal) over a two-year period (Nov 2009–Nov 2011) encompassing 31 storm events. A coupled modeling framework was developed, combining the one-line ShorelineS coastline model with an impact-based dune erosion and recovery module. The model was forced with hourly offshore wave data from the ERA5 reanalysis (corrected for biases in wave height, direction and period using Faro buoy observations) and transformed to the nearshore with the unstructured-grid SnapWave model. In situ LiDAR surveys and satellite-derived shorelines and dune vegetation lines (using Normalized Difference Vegetation Index (NDVI)) were used for calibration and validation. Model results show that longshore sediment transport gradients dominated shoreline change across most of the island, with net erosion in the sediment-starved western sector and net accretion in the eastern sector. Notably, an erosion "hotspot" in the central-west (transects T42–77) experienced intense storm-induced dune retreat, which supplied sand to the adjacent beach and caused local shoreline advance counter to the regional trend. Overwash at the eastern end contributed to the accretion aligned with the regional trend. The dune module's performance was sensitive to its calibration: a higher wave impact coefficient Cs led to greater dune erosion, while coarse sand d50 and low winds markedly reduced modeled post-storm dune recovery. Model validation using satellite-derived shoreline and dune positions employed R² and Pearson correlation metrics, revealing moderate shoreline performance and weaker dune agreement – though general retreat trends were captured across several transects. A correlation analysis indicates that cumulative cross-shore sediment flux (corr = 0.73) and sturm duration (corr = 0.48) exert the strongest control on dune retreat whereas peak storm wave direction and initial berm width had minimal influence. Overall, the coupled model captured the observed pattern of alongshore-varying coastal response – including the emergence of a dune erosion hotspot – highlighting the importance of cross-shore sediment exchanges in barrier island evolution and providing useful insights for coastal management of dune systems.
- Preprint
(4499 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-2384', Anonymous Referee #1, 16 Jul 2025
-
AC1: 'Reply on RC1', Ronaldyn Dabu, 29 Jan 2026
We thank both reviewers for their valuable comments. Below we copied their reviews in black and addressed each point in blue.
Reviewer #1
This study investigated the alongshore variability of shoreline and dune evolution on Culatra Island, a barrier island in southern Portugal, over a two-year period with 31 storm events. A coupled modeling framework was developed to combine the ShorelineS one-line coastline model and a dune erosion/recovery module. The model validation was performed using LiDAR surveys and satellite-derived vegetation lines. Overall, the study is comprehensive, and some findings might be helpful in practice. However, I have several concerns and suggestions regarding the methodology and results of this study as .
Methodology
1) Section 3.2: The temporal resolutions of buoy records and ERA5 datasets are different. How did you compare both? Given the uncertainty and measurement errors in the buoy data, it seems that a to “correct” the original ERA5 data.
We utilized all available buoy observations despite their coarser temporal resolution (typically three-hourly). To align them with the continuous hourly ERA5 data, we matched each buoy record to the nearest ERA5 timestamp. For wave height and period, we applied linear regression-based corrections aimed at minimizing bias and maximizing R². For wave direction, given its sector-dependent bias, we used a 30-degree sector-wise correction approach. This multi-step method resulted in improved alignment between ERA5 and buoy data across all three wave paramete
2) Lines 213 and 214: How did you justify the definition of a storm? Is the number of storms sensitive to the thresholds of the wave height (2.5 m) and the duration (4 hours)?
While a threshold of 3.0 m is commonly used in the literature (Almeida et al., 2012), we conducted a comparative analysis by plotting ERA5 wave heights (y-axis) against Faro buoy observations (x-axis). The scatter plot revealed two distinct trends: one for wave heights below 2.5 m and another above 2.5 m, indicating a shift in behavior and correlation structure around this threshold. Based on this observed breakpoint and its relevance to local morphological response, we selected 2.5 m as a more appropriate threshold for identifying storm events in our study.
Although many studies adopt a 6-hour minimum storm duration to ensure comparability across tide-influenced systems, the 4-hour threshold used here was more appropriate for the local hydrodynamic and geomorp
3) What does RSLR in Equation (3) stand for? What is “K” in Line 248 in Equation (4)? Please make sure the denotations of different variables are consistent in both equations and the text.
RSLR means relative sea level rise (will be included in the R1 version), and K means the transport coefficient (already included in the R0 version).
4) Lines 270-272: The , and it is already provided in the figures.
We appreciate the reviewer’s attention to clarity and brevity. However, we believe the current caption for Figure 5 serves an important explanatory purpose. It does not merely repeat the figure labels but highlights key morphological processes and links the LiDAR profiles to the broader discussion of erosion types and model behavior. This description is intended to help readers interpret the figure without referring back to the main text. We have reviewed the caption for brevity and ensured it avoids unnecessary repetition, but we have retained the key explanatory elements to preserve its standalone clarity.
5) Section 4.4: How did you consider the value of qscal as 0.3? The calibration of Cs and d50 was mentioned in both the Abstract and Conclusions, but the relevant details were only presented in Appendix B. Part of Section 4.4 already presented some results on calibration. It is suggested to move those to the Results and Discussion Section.
We will add a for calibration using different combination of Cs, in the Appendix. Based on previous literature and manual of ShorelineS, qscal can be the calibrating factor for shoreline while Cs affects the dune model.
Results
6) Figure 9 clearly shows that the model output could not capture the net shoreline change, either overestimation or underestimation (and why?). Was the model calibration sufficient for further analysis, or maybe a better model is needed? Given the poor performance of the model, how would you justify the validity of the conclusions drawn based on the model output?
The model captures the general trend such as accretion in the east and erosion in the west.At the ebb delta, the observed erosion is not reproduced but there is a clear minimum in the simulated erosion. As follows from the discussion of Figure 8, the point-by-point correlation coefficient is 0.4, which is not great but in line with typical morphological applications.
7) Figure 9: How did you calculate the BIAS? Is it a total bias or average bias over space?
We used the Average .
8) Lines 397-404: what is the regression model the R2 is measuring? If it is a linear model with an intercept term, R2 is actually equal to the square of the Pearson correlation coefficient, r. Please note that R2 and r are not able to capture the bias in the model prediction. Given the sampling uncertainty in both time and space, it is also suggested to present the values of metrics with a statistical distribution instead of a fixed number. The authors can refer to the article below for more information about the limitations of some commonly used evaluation metrics.
Note that in t, R² is not simply the square of Pearson’s r. Negative R² values occur when the model predictions perform worse than the mean benchmark, while r still captures linear association regardless of model skill.
Reference: “Beyond a fixed number: Investigating uncertainty in popular evaluation metrics of ensemble flood modeling using bootstrapping analysis” (https://doi.org/10.1111/jfr3.)
9) Line 483: The term “dSds” looks confusing. Please elaborate on this denotation.
dSds means longshore and will be elaborated.
10) Section 5.5 and Lines 646-647: To avoid confusion, it is suggested to use lowercase, “r”, to represent the correlation coefficient.
Thank you for this comment, we have made this change.
11) Figures 2, 4 and 7: Please change the color of the cell size labels and transect numbers, for example, it is hard to tell “800 m” and “50 m” from the figure. And add units to the legend of bed elevation. What is the reference level (i.e., 0 m) of bed elevation? What do the contour lines represent in Figure 4? A scale is needed for most maps. Also, please improve the resolution of all the .
We will do
Citation: https://doi.org/10.5194/egusphere-2025-2384-RC1
Citation: https://doi.org/10.5194/egusphere-2025-2384-AC1
-
AC1: 'Reply on RC1', Ronaldyn Dabu, 29 Jan 2026
-
RC2: 'Comment on egusphere-2025-2384', Anonymous Referee #2, 16 Jan 2026
This study employs a numerical modeling approach to simulate shoreline and dune evolution on a barrier island in southern Portugal over a two-year period (2009–2011). The modeling framework is built upon the well-established one-line model ShorelineS, with waves as the primary forcing mechanism. A key contribution is the integration of a dune erosion and recovery module, enabling the investigation of both cross-shore and alongshore morphodynamic responses to storm events. The authors combine bias-corrected ERA5 reanalysis data with in-situ buoy observations and utilize the unstructured spectral wave model SnapWave to generate high-resolution nearshore wave conditions. For model calibration and validation, the study used both LiDAR surveys and satellite-derived shoreline and dune positions extracted using NDVI-based vegetation lines, though this dual-source approach introduces some inherent challenges related to data uncertainty and spatial-temporal resolution. Overall, the study demonstrates the potential of coupled shoreline–dune modeling frameworks to capture alongshore-varying coastal responses, providing valuable insights for coastal modelers. However, certain aspects of the methodology and results presentation require greater clarity and explicit justification to strengthen the manuscript.
Major concerns:
- According to the observation in Figure 9, a clockwise rotation is observed. The modeled shoreline successfully reproduced the rotation but at a much smaller scale. A comparison with original ShorelineS will be welcomed as it has showed good ability in simulating beach rotation and it could be really helpful to investigate how much improvement this study made.
- The dune change needs to be illustrated more explicitly from the source/sink term to volume change and dune toe moving. The methodology lacks clarity regarding which parameters remain constant and which are time-dependent, such as berm width.
- The calibration only uses two LiDAR data in 2009 and 2011 which may not sufficient to calibrate the model. In figure 11, the model only exhibits a long-term trend for both shoreline position and dune position and is unable to capture the seasonal variations. The model was calibrated by LiDAR data and validated through satellite-derived data, cross-validating these two distinct data sources presents significant challenges.
Specific comments:
- Line 157-159: It would be helpful to have more descriptions on PyBeach like what is the input (2D profile?) and how it recognizes the key feature of dune.
- Line 161-162: Can you add the sampling frequency or numbers of available images?
- Line 162: The reference of MNDWI and NDVI are missing.
- Line 180-181: Were ERA5 underestimated over all the time or just storms? This seems inconsistent with the previous paragraph that Faro buoy missing data in some storm events. According to Appendix the bias seems existing for all the period. How two dataset in different sampling frequencies were compared?
- The units of key parameters in section 3.2 are missing, e.g., Hs, Tp.
- Line 199-200: What bathymetry data was used for SnapWave?
- Figure 3: Tide chart could be put in different figure or removed as it is not used as model input? Hs_25m_depth is labeled in the figure but hard to be found in the figure. What is the meaning?
- Line 232-233: More description about the direction would be great. I suppose s is alongshore direction while n represents the shoreline in cross-shore direction.
- Equation 3: This study seems not considering RSLR therefore it is better to remove this term from governing equation or state clearly in the following paragraph.
- Line 241-243: How the separated Dc are reflected in the following calculation? Can you explain more how you did employ them?
- Line 249-250: It might be better to put the description of transport formula you applied before the equation and indicate qscal in the equation if possible.
- Figure 4: Colored lines need to be labeled in legend. And the numbers of transects are not quite clear.
- Similar to previous comment, can you elaborate more basic info about PyBeach? e.g., 2D profile is needed? Apart from that, in transect 36, the location of toe in 2009 and 2011 seems a bit tricky that they both have bumping in front of crest while in 2009 the nearest low point is recognized as toe and in 2011 it is the furthest low point. This can lead to a 60-meter difference of toe position.
- Are "Transect," "Segment," and "Station" referring to the same concept? The terminology requires clarification.
- Figure 5: Segment numbers need to be highlighted. As four subfigures share the same legend, it can be put aside with bigger font for better illustration. This figure could be combined with Figure 4 showing the location of four representative transects. MSL or reference water level can be added to show the location of shoreline.
- Line 312: Some letters with subscript are not displayed correctly.
- Line 319-320: The souce/sink terms with letters are little confusing here. An summation equation would be helpful like: \Sigma_qi = qs+ql+qw (mind the negative and positive symbols)
- Line 324: In Figure 14, why there is an over wash component contributing to shoreline change if qL is not added to the beach?
- Line 325: Is here qs or ql?
- Line 330-331: What wind data was used in this study? It is not stated in the previous data section.
- Section 4.4: How the model calibration was carried out needs to be illustrated explicitly. Only shoreline position in 2011 was used as reference or dune toe position was also used? What evaluation metric is used for optimization? Is Dc here a spatial varied parameter needs to be calibrated or calculated from beach profile data?
- Figure 9: a scale bar for reference is missing.
- Line 383: The term "dune foot/toe" should be consistently used throughout the manuscript to prevent confusion.
- Figure 10 (a): The dates should be labeled more explicitly in the legend and I would recommend to use a clearer legend instead of a screenshot.
- Line 404-405: How are dune toe positions calculated? The governing equation (Eq.3) only contains the shoreline change. Although the dune volume change is calculated in Eq. 7 but how was volume change reflected in toe moving seems missing. Besides that, the dune toe position is acquired based on the same transects as shoreline? The dune lines are showing some concave shapes at some locations in Figure 10(a), does this lead to any problems when extracting dune positions?
- Line 482: Can you check the unit of cumulative changes?
- Line 484 and Line 491-493: Similar to the previous comments, it is little confusing that how overwash affects the shoreline change.
- Figure 14: Can you explain more on this figure that why some transects have both positive and negative dSds bars? Should the cumulative shoreline change look like the same Figure 9(b)?
- Line 518: The definition of duneline needs clarification.
- Line 546-547: In Figure 16 (a), there are some scatters with low qs but have large dune change, could you please explain it more? Should the qs be perfectly matched with dune change according to the model assumption? The overwash and qw terms are relatively small according to previous figures. Can the time step difference between dune/shoreline update and wave calculation lead to this?
- Line 579-580: Are the berm width and slope of each transects updated at each time step?
Citation: https://doi.org/10.5194/egusphere-2025-2384-RC2 -
AC2: 'Reply on RC2', Ronaldyn Dabu, 29 Jan 2026
REVIEWER 2:
This study employs a numerical modeling approach to simulate shoreline and dune evolution on a barrier island in southern Portugal over a two-year period (2009–2011). The modeling framework is built upon the well-established one-line model ShorelineS, with waves as the primary forcing mechanism. A key contribution is the integration of a dune erosion and recovery module, enabling the investigation of both cross-shore and alongshore morphodynamic responses to storm events. The authors combine bias-corrected ERA5 reanalysis data with in-situ buoy observations and utilize the unstructured spectral wave model SnapWave to generate high-resolution nearshore wave conditions. For model calibration and validation, the study used both LiDAR surveys and satellite-derived shoreline and dune positions extracted using NDVI-based vegetation lines, though this dual-source approach introduces some inherent challenges related to data uncertainty and spatial-temporal resolution. Overall, the study demonstrates the potential of coupled shoreline–dune modeling frameworks to capture alongshore-varying coastal responses, providing valuable insights for coastal modelers. However, certain aspects of the methodology and results presentation require greater clarity and explicit justification to strengthen the manuscript.
Major concerns:
- According to the observation in Figure 9, a clockwise rotation is observed. The modeled shoreline successfully reproduced the rotation but at a much smaller scale. A comparison with original ShorelineS will be welcomed as it has showed good ability in simulating beach rotation and it could be really helpful to investigate how much improvement this study made.
We agree that the modeled shoreline rotation differs from the observations. In the present simulations, the longshore component tends to overestimate the rotational signal, while the cross-shore contribution induces shoreline advance along the entire barrier island coastline. This cross-shore advance does not compensate for the overestimated rotation and therefore does not resolve the discrepancy. To fully reflect the processes observed along the barrier island, we quantified the individual contributions of longshore and cross-shore components, as shown in Figure 14. This process-based decomposition allows the relative role of each mechanism to be evaluated without requiring additional model runs.
- The dune change needs to be illustrated more explicitly from the source/sink term to volume change and dune toe moving. The methodology lacks clarity regarding which parameters remain constant and which are time-dependent, such as berm width.
We agree. We will revise the methodology to explicitly describe how source/sink terms are translated into dune volume change and dune toe movement. We will also clearly distinguish constant parameters (e.g., grain size, Cs) from time-dependent parameters (e.g., berm width, runup) in the dune module.
- The calibration only uses two LiDAR data in 2009 and 2011 which may not sufficient to calibrate the model. In figure 11, the model only exhibits a long-term trend for both shoreline position and dune position and is unable to capture the seasonal variations. The model was calibrated by LiDAR data and validated through satellite-derived data, cross-validating these two distinct data sources presents significant challenges.
We agree that the limited number of LiDAR surveys constrains the calibration to long-term trends. We will clarify that the model is not intended to resolve seasonal variability. Satellite-derived data are used for qualitative and trend-based validation rather than strict cross-validation. The uncertainty arising from combining different data sources will be discussed in data sources and model forcings.
Specific comments:
- Line 157-159: It would be helpful to have more descriptions on PyBeach like what is the input (2D profile?) and how it recognizes the key feature of dune.
We will add clearer description stating that PyBeach operates on 2D cross-shore elevation profiles and identifies dune toe and crest using slope, curvature, and machine-learning-based classifiers. Although a more elaborate discussion of PyBeach is already in the paper of Beuzen (2019) as stated in the paragraph.
- Line 161-162: Can you add the sampling frequency or numbers of available images?
We will add the total number of satellite images used and their average temporal sampling frequency.
- Line 162: The reference of MNDWI and NDVI are missing.
The missing references will be added
- Line 180-181: Were ERA5 underestimated over all the time or just storms? This seems inconsistent with the previous paragraph that Faro buoy missing data in some storm events. According to Appendix the bias seems existing for all the period. How two dataset in different sampling frequencies were compared?
We will clarify that ERA5 underestimation occurs over the entire period, with stronger bias during storms where buoy data is available. We will explain that buoy observations were matched to the nearest ERA5 timestamps to account for different sampling frequencies.
- The units of key parameters in section 3.2 are missing, e.g., Hs, Tp.
Units will be added for all parameters.
- Line 199-200: What bathymetry data was used for SnapWave?
We will clarify that SnapWave was forced using regional bathymetry derived from hydrographic surveys and nautical charts, interpolated onto the unstructured grid.
- Figure 3: Tide chart could be put in different figure or removed as it is not used as model input? Hs_25m_depth is labeled in the figure but hard to be found in the figure. What is the meaning?
Tides were used as input both for the wave and shoreline model, Hs at 25 m depth is used to force the dune model. We will explain this more thoroughly in the revised paper.
- Line 232-233: More description about the direction would be great. I suppose s is alongshore direction while n represents the shoreline in cross-shore direction.
We will explicitly state that s denotes the alongshore direction and n denotes the cross-shore (shore-normal) direction.
- Equation 3: This study seems not considering RSLR therefore it is better to remove this term from governing equation or state clearly in the following paragraph.
We agree. The RSLR term will be removed and be explicitly stated as zero.
- Line 241-243: How the separated Dc are reflected in the following calculation? Can you explain more how you did employ them?
We will expand the explanation of how spatially varying active profile heights are applied in the shoreline evolution calculations. Spatially varying values of the active profile height, were used as a calibration parameter in the model. Following the ShorelineS formulation, Dc converts sediment volume change into shoreline displacement through the mass conservation equation. Larger Dc values reduce shoreline movement for a given sediment flux, while smaller Dc values increase shoreline sensitivity. The spatial variation of Dc therefore reflects alongshore differences in profile geometry and dune–beach configuration along the barrier island. We will clarify that these spatially varying Dc values directly control the magnitude of shoreline response to both longshore transport gradients and cross-shore source/sink terms, consistent with the ShorelineS technical reference.
- Line 249-250: It might be better to put the description of transport formula you applied before the equation and indicate qscal in the equation if possible.
We will describe the selected longshore transport formulation before presenting the equation. The transport scaling factor qscal will be explicitly included in the equation, following the ShorelineS formulation. Although a more detailed description of the transport formulation is already explained in the paper of Roelvink et al., 2020, as stated in the paragraph.
- Figure 4: Colored lines need to be labeled in legend. And the numbers of transects are not quite clear.
We will label all colored lines in the legend and increased the size and contrast of transect numbers.
- Similar to previous comment, can you elaborate more basic info about PyBeach? e.g., 2D profile is needed? Apart from that, in transect 36, the location of toe in 2009 and 2011 seems a bit tricky that they both have bumping in front of crest while in 2009 the nearest low point is recognized as toe and in 2011 it is the furthest low point. This can lead to a 60-meter difference of toe position.
We clarified that PyBeach uses 2D cross-shore elevation profiles as input. Dune toe locations are detected based on slope and curvature thresholds. For profiles with multiple local minima, toe identification can vary. We will discuss this uncertainty and its potential impact on dune toe position.
- Are "Transect," "Segment," and "Station" referring to the same concept? The terminology requires clarification.
Yes. All three terms refer to the same alongshore discretization. We will use a single term consistently throughout the manuscript.
- Figure 5: Segment numbers need to be highlighted. As four subfigures share the same legend, it can be put aside with bigger font for better illustration. This figure could be combined with Figure 4 showing the location of four representative transects. MSL or reference water level can be added to show the location of shoreline.
We highlighted segment numbers, enlarge the shared legend, and added mean sea level as a reference water level. The locations of representative transects will be clearly indicated.
- Line 312: Some letters with subscript are not displayed correctly.
This is corrected.
- Line 319-320: The souce/sink terms with letters are little confusing here. An summation equation would be helpful like: \Sigma_qi = qs+ql+qw (mind the negative and positive symbols)
We will add a summation equation defining the total source/sink term. Each component will be clearly defined, including sign conventions.
- Line 324: In Figure 14, why there is an over wash component contributing to shoreline change if qL is not added to the beach?
We clarified that overwash sediment is mainly transported landward. A smaller fraction is deposited at the fronting beach. This fraction is set to 0.2, following the alpha coefficient in the ShorelineS overwash formulation.
- Line 325: Is here qs or ql?
It is ql
- Line 330-331: What wind data was used in this study? It is not stated in the previous data section.
Observed data from Faro.
- Section 4.4: How the model calibration was carried out needs to be illustrated explicitly. Only shoreline position in 2011 was used as reference or dune toe position was also used? What evaluation metric is used for optimization? Is Dc here a spatial varied parameter needs to be calibrated or calculated from beach profile data?
We only calibrated using shoreline positions. We used the coefficient of determination R2 as an evaluation metric. Dc is spatially varied that was calculated from the beach profile data.
- Figure 9: a scale bar for reference is missing.
We added a scale bar
- Line 383: The term "dune foot/toe" should be consistently used throughout the manuscript to prevent confusion.
We now use a consistent term of dune toe.
- Figure 10 (a): The dates should be labeled more explicitly in the legend and I would recommend to use a clearer legend instead of a screenshot.
We now use a clearer label of the dates of the satellite images.
- Line 404-405: How are dune toe positions calculated? The governing equation (Eq.3) only contains the shoreline change. Although the dune volume change is calculated in Eq. 7 but how was volume change reflected in toe moving seems missing. Besides that, the dune toe position is acquired based on the same transects as shoreline? The dune lines are showing some concave shapes at some locations in Figure 10(a), does this lead to any problems when extracting dune positions?
We clarified that dune toe positions are extracted along the same transects as shoreline positions using PyBeach. Dune volume change is translated into toe movement through changes in cross-shore profile geometry. The 1D model then just assumed a constant profile shape that moves depending on the volume change. We will discuss uncertainties related to concave dune shapes.
- Line 482: Can you check the unit of cumulative changes?
Units should be in cubic meters and will be corrected.
- Line 484 and Line 491-493: Similar to the previous comments, it is little confusing that how overwash affects the shoreline change.
We will clarify that overwash mainly acts as a sediment sink from the shoreline system, with only a limited fraction contributing to shoreline change.
- Figure 14: Can you explain more on this figure that why some transects have both positive and negative dSds bars? Should the cumulative shoreline change look like the same Figure 9(b)?
We will clarify that positive and negative bars represent local longshore gradients. The cumulative shoreline change results from integrating these local contributions.
- Line 518: The definition of duneline needs clarification.
We defined the duneline as the alongshore connection of dune toe positions extracted from cross-shore profiles.
- Line 546-547: In Figure 16 (a), there are some scatters with low qs but have large dune change, could you please explain it more? Should the qs be perfectly matched with dune change according to the model assumption? The overwash and qw terms are relatively small according to previous figures. Can the time step difference between dune/shoreline update and wave calculation lead to this?
We will explain that dune change is not controlled by qs alone. Storm sequencing, overwash, and differences in update time steps between shoreline and dune modules can produce large dune changes at low qs.
- Line 579-580: Are the berm width and slope of each transects updated at each time step?
Yes. Berm width and slope are dynamically updated at each time step following shoreline and dune toe evolution.
Citation: https://doi.org/10.5194/egusphere-2025-2384-AC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 894 | 135 | 38 | 1,067 | 27 | 39 |
- HTML: 894
- PDF: 135
- XML: 38
- Total: 1,067
- BibTeX: 27
- EndNote: 39
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This study investigated the alongshore variability of shoreline and dune evolution on Culatra Island, a barrier island in southern Portugal, over a two-year period with 31 storm events. A coupled modeling framework was developed to combine the ShorelineS one-line coastline model and a dune erosion/recovery module. The model validation was performed using LiDAR surveys and satellite-derived vegetation lines. Overall, the study is comprehensive, and some findings might be helpful in practice. However, I have several concerns and suggestions regarding the methodology and results of this study as follows.
Methodology
1) Section 3.2: The temporal resolutions of buoy records and ERA5 datasets are different. How did you compare both? Given the uncertainty and measurement errors in the buoy data, it seems that a linear correction is good enough to “correct” the original ERA5 data.
2) Lines 213 and 214: How did you justify the definition of a storm? Is the number of storms sensitive to the thresholds of the wave height (2.5 m) and the duration (4 hours)?
3) What does RSLR in Equation (3) stand for? What is “K” in Line 248 in Equation (4)? Please make sure the denotations of different variables are consistent in both equations and the text.
4) Lines 270-272: The description about Figure 5 is redundant, and it is already provided in the figures.
5) Section 4.4: How did you consider the value of qscal as 0.3? The calibration of Cs and d50 was mentioned in both the Abstract and Conclusions, but the relevant details were only presented in Appendix B. Part of Section 4.4 already presented some results on calibration. It is suggested to move those to the Results and Discussion Section.
Results
6) Figure 9 clearly shows that the model output could not capture the net shoreline change, either overestimation or underestimation (and why?). Was the model calibration sufficient for further analysis, or maybe a better model is needed? Given the poor performance of the model, how would you justify the validity of the conclusions drawn based on the model output?
7) Figure 9: How did you calculate the BIAS? Is it a total bias or average bias over space?
8) Lines 397-404: what is the regression model the R2 is measuring? If it is a linear model with an intercept term, R2 is actually equal to the square of the Pearson correlation coefficient, r. Please note that R2 and r are not able to capture the bias in the model prediction. Given the sampling uncertainty in both time and space, it is also suggested to present the values of metrics with a statistical distribution instead of a fixed number. The authors can refer to the article below for more information about the limitations of some commonly used evaluation metrics.
Reference: “Beyond a fixed number: Investigating uncertainty in popular evaluation metrics of ensemble flood modeling using bootstrapping analysis” (https://doi.org/10.1111/jfr3.12982)
9) Line 483: The term “dSds” looks confusing. Please elaborate on this denotation.
10) Section 5.5 and Lines 646-647: To avoid confusion, it is suggested to use lowercase, “r”, to represent the correlation coefficient.
11) Figures 2, 4 and 7: Please change the color of the cell size labels and transect numbers, for example, it is hard to tell “800 m” and “50 m” from the figure. And add units to the legend of bed elevation. What is the reference level (i.e., 0 m) of bed elevation? What do the contour lines represent in Figure 4? A scale is needed for most maps. Also, please improve the resolution of all the figures.