Mediterranean UNESCO World Heritage at risk from coastal flooding and erosion due to sea-level rise

UNESCO World Heritage sites (WHS) located in coastal areas are increasingly at risk from coastal hazards due to sea-level rise. In this study, we assess Mediterranean cultural WHS at risk from coastal flooding and erosion under four sea-level rise scenarios until 2100. Based on the analysis of spatially explicit WHS data, we develop an index-based approach that allows for ranking WHS at risk from both coastal hazards. Here we show that of 49 cultural WHS located in low-lying coastal areas of the Mediterranean, 37 are at risk from a 100-year flood and 42 from coastal erosion, already today. Until 2100, flood risk may increase by 50% and erosion risk by 13% across the region, with considerably higher increases at individual WHS. Our results provide a first-order assessment of where adaptation is most urgently needed and can support policymakers in steering local-scale research to devise suitable adaptation strategies for each WHS.

This paper has more to say about the problem than about the solution-a criticism that could be equally leveled at almost every similar study of the past decade, including my own. However, I think the authors could say more in the discussion about others around the world who have tried to confront this problem, and offer more concrete solutions. In particular, they might look for examples in the work of Scottish Coastal Heritage at Risk program, and similar efforts in Ireland and England. The United States' National Park Service Cultural Resources Climate Change Strategy might also be useful. I am sure there are other worthy efforts around the world, but these are the ones I am most familiar with.
In sum, the authors have presented a concise, effective study of the impacts of sea level rise on World Heritage Sites around the Mediterranean. The methodology is thorough, but simple enough to be reproduced in many other parts of the world. I hope the next decade will produce similar research focused not only on evaluating the magnitude of the problem faced by cultural heritage, but also in documenting efforts to mitigate those problems.

--Leslie Reeder-Myers
Reviewer #3 (Remarks to the Author): The paper proposed by the authors deals with an interesting aspect, the impact of coastal hazards on cultural heritage, which is of interest to the physical science community as well the archaeological community.
The idea of doing estimations at Mediterranean level is remarkable but the paper fails to convince the reader that the analyses done are sound and robust. Although at a qualitative level proposes some interesting hypotheses, at a quantitative level it is full of assumptions that shade a light of inaccuracy on the work.

FLOOD RISK
The paper in its current form refers to another paper that was under review: Luckily I discovered that the paper has just been published as https://www.nature.com/articles/sdata201844#f3 I went to consult that paper and I can take that for flood risk it produced an innovative dataset of return period levels for surges.
The main problem for this type of work is the fact that it relies heavily on the SRTM dataset which has the consensus view that it has a minimum vertical accuracy of 16 m absolute error at 90% confidence (Root Mean Square Error (RMSE) of 9.73 m) world-wide.
The paper draws some strong conclusion for sites located at deltaic areas like the Po and the Rhone Delta, which have an elevation of few metres abobe MSL when are not below MSL. I have the feeling that with this type of inaccuracy of the dataset, the inundation model has not the capability of distinguish real flooding from artefacts. Moreover, the morphological data on the shoreline does not include anything about the level of protection, neither the real elevation of dykes and embankments, nor their characteristics.
Indeed the authors admit that :" We do not account for existing flood protection measures in our analysis due to a lack of consistent region-wide data." However, for some regions like the Po area and its southern coastline these information could have been obtained from local knowledge data and at least used to carry out a sensitivity analysis on the results. This coastline is heavily protected and it seems unlikely that the current level of defence would not be adapted to rising sea-level.

EROSION RISK
If the flood risk, after proper evidence of reliability in data-rich regions, may seem acceptable, the erosion risk makes little sense. The fact that risk is determined by a site's distance from the coast is a simplistic viewpoint.
A strong role in this evaluation is played by the "sediment supply" parameter, making reference to This is a broken link and in any case, in all the DIVA model's output I have never seen meaningful information on sediment supply.
Moreover the statement "The sediment supply index includes qualitative information on sediment supply; a high index (maximum of 5) represents low sediment supply and a low index (minimum of 1) represents high sediment supply" confirms my worries. It is a totally subjective index.
The paper as it is not acceptable and scientifically sound. I propose to the authors to re-write the paper considering:

1)
To eliminate the Erosion Index

2)
To produce evidence that the Flooding Index is reliable by taking a data-rich example and use a Lidar dataset to compare the output of an SRTM approach with a detailed Lidar approach and carry out a sensitivity analysys for the flood index. to sea-level rise" We would like to thank all reviewers and the editor for their valuable and constructive comments. We have carefully considered all comments and have substantially revised the manuscript in response to the points that the reviewers have raised. Major changes include: -repeating the model runs, using regional SLR projections, including 10-year time steps to account for the temporal evolution of the risk indices and for changes in the coastline due to permanent inundation induced by SLR -providing further literature-based justification for the use of the erosion risk index and the weighting of its individual parameters -an extended discussion on the relevance of our results for policymakers on regional scales -a concise discussion on the limitations of this study (DEM, coastal defences) Below you can find the responses to each point raised by the reviewers (replies in red italics).

Reviewer #1
I would suggest that the paper could be appropriate for publication with the following major changes: 1. A clearer explanation of the theoretical basis for the weightings used in the indices 2. A justification for the linear combination of erosion risk and flood risk, or else dropping the linear combination and presenting these separately 3. A clear explanation of how the authors' expect a planner to make use of the indices presented. 2.) Moreover, the rationale for linearly combining the flood risk index and the erosion risk index is unclear, as is the rationale for the particular weightings used in the erosion risk index.
We agree with the comment of the reviewer that the combined index was not meaningful in supporting decision-makers. We have therefore removed the linear combination of the two indices.
Regarding the weightings of the erosion risk index: these are based on the current literature that assesses coastal erosion [6][7][8][9][10][11][12] and cultural heritage at risk from coastal erosion 13-15 on regional scales and specifically on the weightings used in Reeder-Myers et al. (2015) 13  3.) Conceivably, since each site can at least be compared to itself at a different point in time, the evolution of the flood risk index over time could highlight when the hazard at different sites becomes of concern, but the authors' use of a single end-of-century time point means it cannot happen based on current results.
The comparison of different scenarios might also help indicate the value of global climate mitigation for WHS protection, but since the scenarios used are the low end of the AR5 likely range for RCP 2.6, the high end of the AR5 likely range for RCP 8.5, and a third, higher scenario, the analysis in its present form does not lend itself to this comparison.
Indeed, from the current analysis it is unclear how much of the flood risk is due to sea-level rise, and how much exists under current conditions (i.e,. 0.0 m AMSL). , and it's unclear why this uncertainty cannot simply be taken into account in the scenario definition. Other sources of uncertainty also are present, and yet are taken into account; and neglecting non-dynamic drivers of differences between global-mean sea-level change and regional sea level change (e.g., static equilibrium effects) seems unjustified. Per Kopp et al., 2014, the median ratio for RCP 8.5 in 2100 of climaticallydriven (i.e., excluding GIA and tectonics) RSL rise to GMSL rise in the Mediterranean is about 0.8-0.9.
Indeed, the authors use as one of their sources the sea-level rise projections Jackson and Jevrejeva (2016). The main point of this paper is the production of regional sea-level projections. If these projections are not to the authors' liking, others (e.g., Kopp et al., 2014Kopp et al., , 2017) also provide regionalized, probabilistic projections of RSL change around the world.
As an alternative approach, the authors might instead look at the flood index at different arbitrary but evenly spaced SLR levels (e.g., 0 to 2 m in 0.25 m spacing). This separates the question of flood risk from the question of sea-level rise timing and emissions sensitivity, which can then be addressed in a separable analysis. For example, the authors could for each site identify the RSL rise at which the site crosses a given flood risk threshold, and then draw upon the Jackson & Jevrejeva or Kopp et al. projections close to that site to determine the range of when that threshold might be crossed under different scenarios. This last identification could inform planners about when it would become necessary to prioritize coastal flood adaptation at different sites.
We have repeated our analysis using the regionalized SLR projections by Kopp

Reviewer #2
This paper presents an important contribution to recent studies that focus on the impacts of climate change on archaeology and cultural heritage around the world. The authors look at UNESCO World Heritage Sites around the Mediterranean, and examine the risk of flood and erosion based on different sea level rise scenarios. They use publically available data, modified as necessary to more accurately reflect real world conditions, to determine that around three quarters of World Heritage Sites in this region are threatened by sea level rise, even under the most conservative estimates. They present their data and methodology in a clear, reproducible manner. 16.) This type of analysis has become common in recent years, and the authors place their study in this larger context of research in Europe and around the world. Instead of focusing on all archaeological sites within a small region, which is a common approach, this study focuses on some of the highest profile cultural heritage sites in the world. In many ways, this is important. By highlighting the damage that will certainly occur at these iconic places, the authors draw attention to the dire condition of all cultural heritage in the coastal zone over the coming centuries. Moreover, such sites are more likely to receive the attention (i.e. funding) they need to mitigate those risks. That said, it is important to note that the vast majority of cultural heritage occurs in far less spectacular locations, and that the bulk of archaeological sites will be simply lost, along with the information and connections to the past that are contained in them. An evaluation of that loss clearly goes beyond the scope of the current paper, but the authors could point out that both the more humble and the more spectacular sites are important parts of human history.
We thank the reviewer for raising this point. In sum, the authors have presented a concise, effective study of the impacts of sea level rise on World Heritage Sites around the Mediterranean. The methodology is thorough, but simple enough to be reproduced in many other parts of the world. I hope the next decade will produce similar research focused not only on evaluating the magnitude of the problem faced by cultural heritage, but also in documenting efforts to mitigate those problems.

--Leslie Reeder-Myers
We would like to thank the reviewer for her positive and constructive comments and we are hopeful that we have responded to her comments in a satisfactory manner.

Reviewer #3
The paper proposed by the authors deals with an interesting aspect, the impact of coastal hazards on cultural heritage, which is of interest to the physical science community as well the archaeological community.
The idea of doing estimations at Mediterranean level is remarkable but the paper fails to convince the reader that the analyses done are sound and robust. Although at a qualitative level proposes some interesting hypotheses, at a quantitative level it is full of assumptions that shade a light of inaccuracy on the work.

We have carefully considered the reviewer's suggestions and have conducted additional analysis to
address the points raised. We have also substantially rewritten large parts of the manuscript to provide additional information regarding the issues of concern.
FLOOD RISK 18.) The paper in its current form refers to another paper that was under review: Wolff, C. et al. A Mediterranean coastal database for assessing the impacts of sea-level rise and associated hazards (under review).
Luckily I discovered that the paper has just been published as https://www.nature.com/articles/sdata201844#f3 I went to consult that paper and I can take that for flood risk it produced an innovative dataset of return period levels for surges.
The main problem for this type of work is the fact that it relies heavily on the SRTM dataset which has the consensus view that it has a minimum vertical accuracy of 16 m absolute error at 90% confidence (Root Mean Square Error (RMSE) of 9.73 m) world-wide.
The paper draws some strong conclusion for sites located at deltaic areas like the Po and the Rhone Delta, which have an elevation of few metres above MSL when are not below MSL. I have the feeling that with this type of inaccuracy of the dataset, the inundation model has not the capability of distinguish real flooding from artefacts.

The reviewer is correct in pointing out the limitations of the SRTM, which is also a point raised by
Reviewer #1 (comment 6). We are aware of the limitations of the SRTM data but, to our knowledge, there is currently no other region-wide DEM available with better characteristics. At the same time the SRTM DEM has been extensively used in large-scale flood modelling 16,[19][20][21][22][23] . Indeed, the DEM has an absolute vertical error of 16 m. Importantly, the relative error is considerably smaller (<10 m), while the absolute vertical error has been shown to be smaller in coastal areas 24,25 . A recent study assessing the limitations of the SRTM data for coastal vulnerability assessments 26  19.) Moreover, the morphological data on the shoreline does not include anything about the level of protection, neither the real elevation of dykes and embankments, nor their characteristics.
Indeed the authors admit that :" We do not account for existing flood protection measures in our analysis due to a lack of consistent region-wide data." However, for some regions like the Po area and its southern coastline these information could have been obtained from local knowledge data and at least used to carry out a sensitivity analysis on the results. This coastline is heavily protected and it seems unlikely that the current level of defence would not be adapted to rising sea-level. This is a broken link and in any case, in all the DIVA model's output I have never seen meaningful information on sediment supply.
Moreover the statement "The sediment supply index includes qualitative information on sediment supply; a high index (maximum of 5) represents low sediment supply and a low index (minimum of 1) represents high sediment supply" confirms my worries. It is a totally subjective index.
We have carefully considered the reviewer's comment. As  Nevertheless, index-based approaches assessing the risk of coastal erosion are well-established in current research [6][7][8][9][10][11][12]40 , with some studies explicitly analysing cultural heritage at risk from coastal erosion [13][14][15] . The risk index allows us to assess WHS without attaching monetary value to them. This approach is particularly suitable for first-order assessments on regional scale to support adaptation planning 10,11,40  The paper as it is not acceptable and scientifically sound. I propose to the authors to re-write the paper considering: 1) To eliminate the Erosion Index 2) To produce evidence that the Flooding Index is reliable by taking a data-rich example and use a Lidar dataset to compare the output of an SRTM approach with a detailed Lidar approach and carry out a sensitivity analysis for the flood index.
Dear authors, I find the manuscript much better than before.
Regarding your statement in the rebuttal letter which I copy below: "As sediment supply plays an important role for coastal erosion in the Mediterranean5,32,33, we use a newly created dataset of mean monthly Total Suspended Matter (TSM) concentration to represent local sediment availability34. TSM is a measure of the turbidity of the water in coastal locations and has been produced based on satellite imagery as part of the GlobColour project35. We spatially join the TSM data to the MCD and subsequently attribute TSM to each WHS. As sediment supply mainly plays a role in calm waters where it can be deposited5, we exclude TSM from the erosion risk index at WHS in rocky locations. Please see lines 393-399, 406-407 and 412-414 for further details." I am still dubious about using as proxy for sediment supply a value of suspended sediment (e.g. turbidity), obtained from satellite measurements. What really counts on counteracting coastal erosion is river bedload at the mouhts (generally sand), while what you obtain from satellite is generally fine suspended sediment, which is dispersed offshore.
Maybe it would be wise to make a statement that you are aware of the limits of this dataset of turbidity data. What really lacks is a global dataset of bedload sediment transport for rivers. Maybe the reference list was not updated, I suggested to do so. In any case I appreciate your awareness of the problem of sediment supply.
Reviewer #4 (Remarks to the Author): I have been asked to examine the authors' revisions and responses to Review #1's comments and concerns regarding the use of global sea level rise scenarios in a local contact.
Reviewer #1 suggested that it was inappropriate to use global sea level projections with some sort of local (presumably GIA) correction. I have not seen the original manuscript, but I would agree. The revised manuscript has addressed this concern by re-doing their analysis with the Kopp et al. (2017) probabilistic projections which are provided at tide gauge sites and on a 2 deg by 2 deg grid. I believe that the modifications that the authors have made sufficiently address Reviewer #1's concerns on this matter.
In addition to focusing on this specific aspect of the manuscript, I have a few additional minor comments below.
It is worth noting though that work has shown that future 100-yr flood heights are biased low when you consider that sea level is rising at an uncertain rate (Buchanan et al., 2016, Climatic Change). While the Mediterranean was not explicitly discussed in , it would be worthwhile for this paper to be cited as it will be important for site-specific adaption efforts.
Line specific comments: Line 81: "HE" has not yet been defined.
Line 113: I think you are missing some words in this sentence. I don't follow it. Figure 4b: this is the change in erosion risk index? Please make this more explicit.