Assessing the Impact of Physical and Anthropogenic Environmental Factors in Determining the Habitat Suitability of Seagrass Ecosystems
Round 1
Reviewer 1 Report
Review of Hastings et al. for Sustainability
Main comments
Overview:
This study developed a species distribution model (SDM) for predicting the distribution (or potential restoration area) for two seagrass species, Zostera marina and Z. noltii in the Republic of Ireland. Environmental and anthropogenic variables were used with the maximum entropy method to develop a suitability metric and maps for potential distribution. Physical variables were found to be more important than anthropogenic variables at the national scale (the scale of the SDM). The study also utilized 53 field surveys to assess a different suite of physical and anthropogenic variables and “pollutions” at a fine geographic scale.
This is an interesting study that presents a useful method for predicting seagrass extent or potential restoration areas on a broad scale when mapping is lacking. Seagrass distribution changes frequently and is difficult to map directly, so this is a useful tool. This study will be useful for the seagrass, blue carbon, and marine restoration communities, but several clarifications, and a few potentially large issues need to be addressed before the study can be published, and the writing could be made more concise and directly relevant to the study, especially in the introduction and discussion. The largest issue of the paper is surrounding the field surveys and the conclusions drawn from them, and I suggest this aspect of the study be removed entirely to improve the quality of the manuscript and the appropriateness of the conclusions drawn.
Title:
The title refers to ‘blue carbon’ but the study is about species distribution models for better describing seagrass distribution OR potential restoration areas. i.e. a second step is required after modeling to determine if seagrass is actually there, or why seagrass isn’t there (other limitations to seagrass survival and/or recruitment). The modeling itself does not result in maps of current distribution, just niche suitability, which is still useful. The title should be changed to better reflect what the study directly measures (which does not include carbon). How about "Modelling habitat suitability of seagrass ecosystems: a case study from the Republic of Ireland"?
Field surveys:
I’m not sure these field surveys should be included in the paper at all. They seem to weaken the overall quality of the study and don’t appear to add any useful findings. The interpretation and conclusions of the field surveys seems to go far beyond the actual data that was collected. The purpose of the 53 field surveys is unclear, mainly due to the definition of seagrass “presence” used in the study, which is incorrect. Seagrass present at the site that is “washed ashore” is not the same as observing live seagrass beds, but rather it should be described as “seagrass wrack”, which could be imported from distances near or far, and does not suggest living seagrass presence at the site. The study claims that 4 “new” seagrass sites were discovered, yet these are based on the presence of wrack, so should not be considered seagrass sites.
Methods surrounding the field surveys are vague, making assessment by the reader difficult. It is not clear how or why the sites were chosen, what methods were undertaken (e.g. transects? Quadrat measurements?), or whether the model output was compared to seagrass observations (Did you expect to find seagrass at the sites? Do findings agree with your modelled suitability distribution?).
The metrics recorded in the field survey are negatively biased and should be re-categorized to avoid this. For example, bioturbators should be called bioturbators, not “pests”. Macroalgal wrack should also be called what it is, as it is naturally deposited on coasts and beaches, and therefore should not be lumped in the “pollution” category, which suggests direct human impacts. I suggest re-categorizing (see specific comments).
Species Distribution Modelling:
-Caveats surrounding applying this model to determine current distribution and/or targeting sites for restoration are not critically discussed (see specific comments below). For example, it should be discussed that sites deemed “suitable” that do not currently have seagrass present should be critically evaluated to determine what is preventing natural seagrass recruitment and/or survival before restoration projects are given the green light to avoid restoration failure.
-An additional or substitute figure for the suitability maps would be useful to show areas that are and are not useful (based on a cut-off value of what is considered “suitable” – perhaps with levels – 0.9 and up is highly suitable, 0.75 and up is moderately suitable, and over 0.5 is likely suitable or something like that.
-More explanation on the meaning and interpretation of “percent contribution” and “permutation importance” are needed to help the reader understand why both metrics are presented, yet permutation importance is used as the final indicator of which variables are deemed more important for the SDM.
-Not enough info on methods and details of the species extent and physical variables used in the model. See specific comments, but for example, how were species distributions mapped previously and what year are those maps from? What is the resolution of bathymetry and other environmental data? What exactly are eco risk and eco status?
-No ground-truthing was used to test whether model predicted locations of new seagrass beds (though this is the claim – that the distribution is actually 4x that of current mapped extent). This is ok, but should be clearly stated.
Specific comments:
Abstract
Line 19: word choice “…or to identify suitable sites…” rather than “inform”
Line 20: briefly state why anthropogenic variables should be included (why is this an important aspect of your study that others have overlooked?).
Line 23: most readers won’t know what MaxEnt is. Re-phrase “Using the Maximum Entropy method for developing the SDM, …”
Line 26: Do you really mean “high standard deviation of SST”? A high variability in SST was important?
Line 30: It seems odd to put natural accumulation of wrack under the category of "pollution" with human-related variables of boat traffic and litter.
Line 30: "appeared to limit the distribution in county Cork." - Because you found few to no living seagrass sites in your survey, this statement is a speculation. It is not possible to say that these are limiting the distribution when you didn't survey sites with healthy Zostera meadows (which could have also had these stressors, but survived).
Line 31: Were seagrass beds expected at these survey sites based on the SDM?
Line 32-33: Were results of the SDM ground-truthed at all? Also, be cautious in wording about “suitable” areas with no seagrass being used for restoration (there could be reasons why seagrass isn’t already there).
Introduction
Line 47: “are” should be replaced by “is”
Line 49: “satisfy” should be “satisfies”
Line 52: “implementing” should be “incorporating”
Line 55-58: sentence tenses odd and “bioindicators…” phrasing doesn’t fit grammatically
Line 48-67: Can these two paragraphs be combined in shortened? Each paragraph takes a separate approach to justifying the need for better mapping, but seem disjointed when presented separately.
Line 68-69: for studying “all” taxa? This seems extremely broad. For studying a “variety” of taxa?
Lines 76-77: Gradients and differences of what? Bathymetry? Salinity?
Line 83: “they are not the sole driver” – isn’t this what you are testing? Rephrase to “they may not be the sole driver”
Lines 84-96: Sentences in this paragraph seem very repetitive. Can you go more into the mechanisms and examples of how anthropogenic drivers may limit seagrass distribution, rather than restating that anthropogenic variables aren’t usually included in SDMs? The latter can be stated one time with multiple references.
Lines 107-109: I don’t think questions 3 and 4 are answered in this study. I would remove these.
Methods:
Lines 125-129: What does “verified” mean in this context? Please explain. How were the seagrass distribution areas mapped/estimated? What year are they from? Were they measured via satellite and ground trothed? Mapped manually?
Figure 1: Geographic labels on your maps would be helpful (“Republic of Ireland” and “county Cork”). A box around Cork in map “a” would also be helpful.
Lines 140-141: It seems odd to convert polygons to points, considering points cover only a small area compared to the potentially large areas covered by polygons. Was seagrass area lost due to conversion of polygons to points? Or were smaller polygons created to overlap the spatial coverage of larger polygons to ensure points covered the majority of the seagrass extent? Please clarify how you were able to maintain true area when converting polygons to points.
Line 146: Much more detail is needed about the field surveys. Please briefly describe this “predominant method” used in Ireland for readers who are unfamiliar. How much area was covered at each site? How were sites chosen? Were these sites compared to the model predictions later on? Were transects used? Quadrats? Did you measure density? How do you define “presence” of seagrass? Much more detail is needed here.
Lines 147-148: sentence structure - "Following the recommendations of the EPA, surveys took place..."
Line 151: Why was 1 km-resolution chosen? Did this match many of the original spatial data layers?
Line 156-157: It bioturbators are generally not considered “pests”, as this gives a negative (biased) connotation to the fauna naturally found in the sediment. Similarly, “seaweed piling” is usually naturally accumulated macroalgal wrack, not a result of human “pollution”. More appropriate categories would be substrate, exposure, pollution, macroalgal wrack, and infauna.
Line 159-160: “exposure and substrate” – details? How were exposure and substrate categorized and determined in the field?
Line 170: “ascertain” should be changed to “predict” or “estimate” since you did not confirm your model predictions of distribution.
Table 1:
-Each statement in the “Justification” column should have references associated with it.
-Methodology column – could this column be better titled as “Spatial Data Pre-Processing” or something similar?
-More information is needed for many of the layers either here or in the text - e.g. what was the original resolution of bathymetry and other data layers? (Briefly) What data collection methods were used to create the layers? What year were the spatial data generated? This info is key for the reader to understand how the environmental variables relate to spatial and temporal scales as compared to the species distribution data and suitability scores generated based on 1 km resolution.
-“WFD” – what does this stand for?
-“Eco status” and “risk status” – it is not clear from the description what these mean or what data were used to generate them.
- "Exposure" usually refers to wave exposure, but it sounds like this is more of a categorical assessment based on anthropogenic risk? Please clarify what this is, as well as Eco Status and Risk Status. How can you get a physical predictor by combining two anthropogenic predictors?
Line 183 – reference?
Lines 184-185: This describes your "suitability" indicator, correct? What is your cut-off for what is considered "suitable" habitat? Is it >0.5? Higher? This should be clearly stated, then applied to specify estimated distribution area in a figure in addition to figure 2 to show maps of "suitable" area (rather than only a range of suitability from 0-1, which is what is shown now). It is difficult to tell from the color gradient how much of figure 2 area is actually “suitable” (e.g. over 0.5). A new set of maps (similar in extent to Figure 2) showing binned levels of suitability would be very useful for showing where suitable areas are, and how they scored compared to one another. For example, a “very high suitability” category for >0.9, a “high suitability” category for 0.75-0.89, and a “moderate suitability” category for >0.5-0.74 (each symbolized in the maps) to show areas of highest priority in Ireland for further investigation (i.e. for searching for unmapped beds, or assessing the sites’ potential for restoration).
Lines 189-192: Sentence long and a bit confusing. Please reword/clarify.
Line 188: Why use all three methods for evaluation? When they give differing results (table 2), which value is more important? Can you briefly describe how to assess these values compared to one another to ease the reader's interpretation of the data (e.g. Table 2). In the results, it seems percent contribution is essentially entirely ignored and permutation importance is used as the default indicator to determine the ranking of important drivers. Why?
Line 203: Again, why was 1km resolution used? Why not finer? Can you provide a justification for this earlier in the methods?
Results
Line 218: Why focus on this instead of percent contribution? Dredge fishing and eco status each have ~10% contribution – is this not important if permutation importance percentages don't reflect these? Please explain how each of these parameters (percent contr vs permutation importance) is assessed, especially when they contradict one another.
Line 223: If there is a “positive” response to bathymetry between 3-10 m, wouldn’t that mean either 3 or 10 would be the ideal bathymetry, not 5 m? Please clarify.
Line 225: “is” should be changed to “was”
Line 228: This seems like an odd/unexpected result. It is counterintuitive to see that a smaller variability in temperature would be negatively associated with seagrass distribution. You do not directly address this result in the discussion. Shouldn’t areas with fewer temperature fluctuations (i.e. less extremes of hot and cold temperature stress) be more suitable for seagrass? Can you speculate or find literature to help explain this result in the discussion (I also commented on below in the discussion).
Line 233: “is” should be changed to “was”
Figure 2
- These maps show a range of suitability - which areas are considered "suitable" vs not? Though it is useful to see the gradient of 1 to 0 for the color symbology, as-is it cannot be utilized to identify potential distribution. An additional figure showing suitability (for the same panel areas for each species) would be useful, with estimates of areal extent of potential distribution for each.
- Can you add labels to the inset boxes and matching zoomed in panels (e.g. "A" on the inside of the northernmost box, with a matching "A" in the top left zoomed panel), also "Ireland" to main map.
-The resolution looks a little blurry for the color and text - can this be improved (i.e. is it related to figure processing) or is this related to the resolution of the modelling?
- The caption should allow the figure to stand alone from the main text. Please expand on this caption to briefly describe how the figure was generated, e.g. "Habitat suitability... based on physical and anthropogenic variables assessed through MaxEnt"
Section 3.2 Field surveys
- Seagrass that is detached or "washed ashore" should be referred to as “seagrass wrack” and should not be confused with living seagrass beds. It does not indicate presence of growing seagrass, and without further investigation it is very difficult, if not impossible, to determine where the seagrass washed in from. Seagrass wrack should not ever be used as a proxy for seagrass “presence”.
-What is a "sample" in this context? Is it a ramet of detached seagrass? This is simply wrack, and does not give a proxy of proximity. Seagrass wrack can be transported over very long distances in a short amount of time. It is only appropriate to report seagrass presence where seagrass was actually seen growing, anchored and alive in the sediments. If no seagrass was found in the surveys (which is how it sounds), the aspects of the study related to the field surveys should not be included, because sites that represent both seagrass presence and absence were not sampled, and therefore links between measured variables and seagrass suitability cannot be made. This section should either be entirely removed (along with methods and discussion related to field surveys) or be re-written to clearly reflect that these surveys are qualitatively informative only (and information on “samples” is not useful). The surveys could be briefly mentioned in that they can be used to suggest potential relationships between the variables observed (e.g. macroalgal wrack presence and lack of living seagrass) that should be investigated in the future, but cannot be used to make conclusions from this study (and this should be clearly and explicitly reflected in the discussion and abstract). This would be a much stronger paper without including the field surveys and I highly suggest you remove this content.
Table 2: Again, why is permutation importance discussed as the metric that determines driver importance, while percent contribution is largely ignored?
Figures 3 and 4. Move to supplementary material. As things are, there is not much in the results about the response curves, and no information on how they are generated or what they are based on. This information should be added in the text, and I think both response curve figures are more appropriate in the supplementary materials if they will not be discussed in more detail. You mention the 3-10m optimal bathymetry range for Z. marina, but it is very hard to distinguish that from this figure because individual panels are quite small.
Line 263: Very vague first sentence. Can this be made more informative?
Lines 263-267: Again, (if not completely removed) this section should be altered based on comments in the methods. Seaweed wrack is not pollution. Bioturbators are not "pests".
Line 266 – “half deemed at risk” – what does this mean? What is it based on?
Lines 269-270: Avoid referring to figures in this way. "It can be seen from Figure 6a" - instead just state "Seaweed pilings were the...(Figure 6a)."
Figure 5: very misleading if seagrass “presence” referred to seagrass wrack observed at the site…
Discussion
General:
It is important to discuss the limitations of your study and your methods -
-Why didn't you "ground truth" any of the predictions from your model to see if seagrass was present where your suitability scores were high?
-In the intro, you discuss the applicability of the SDM for identifying current distribution/range, or identifying areas fit for restoration if the current range doesn't actually cover those suitable areas. Isn't this risky? If seagrass isn't present in areas deemed "suitable" by the model, perhaps there are other factors limiting seagrass distribution that should be assessed to ensure restoration success. Restoration success may be limited by factors not measured in this study, not mapped at relevant scales, or for which there is no data. Survival and recruitment may be limited by seed availability (proximity and/or connectivity to other meadows) or other factors. These should be clearly addressed throughout the paper as a caution. Areas deemed "suitable" by your model may in reality be risky sites for restoration without further investigation. It begs the question why isn't seagrass there already?
-Though you identified a range of suitability levels, some "suitable" areas may not actually have seagrass. What else could be limiting seagrass establishment or survival in these areas? You mention macroalgal wrack, but there are many potential factors that you didn't model, or as you touch on can't be modelled at broad scales, that could impact recruitment and survival. Before undertaking large restoration projects in "suitable" areas, what other factors would be assessed or tested in addition to macroalgal wrack presence to ensure success?
-Some of the discussion text that is very broad compared to the study focused could be cut to discuss in better depth the limitations listed above.
Line 279: Isn't your aim to determine which are most important, not just which type? I recommend rewording: "...identify key physical and anthropogenic environmental drivers..."
Line 284: lots of jargon. Should this be explained in the methods?
Line 285: Are these values reported in the results? If not, add them.
287-292: It seems a lot of this should be moved to the methods and results before discussing it here.
299-324: It seems like the focus of these paragraphs is other research, with mentions of your results as sort of a side note. Can these paragraphs be changed to better highlight the importance of your research in these contexts?
344-349: This seems a bit off topic (beyond the scope of this study), but perhaps could be mentioned just briefly how SST relates to growth and blue carbon (with appropriate references).
Line 325: “it is clear…” – strong wording, please rephrase.
Lines 325-340: I recommend deleting this paragraph. Unattached seagrass (i.e wrack) does not indicate presence nearby. It being “unlikely” to have traveled (lines 329-330) is purely speculation, and not supported by data on seagrass wrack, which can drift great distances. Large masses of seagrass wrack can accumulate far from meadows due to currents and waves, water circulation direction, fetch, and so on.
Line 357: remove “successfully”
Line 359: This seems very odd. Can you explain or speculate on why? Is there anything in the literature to support that more variability in temperature is somehow beneficial for seagrass growth/survival? I would expect the opposite – extreme highs and lows would be associated with intolerance due to heat and cold stress.
Figure 6: Again, I think the study would be better without the inclusion of the field surveys. Pie charts don’t seem appropriate for comparing “pollution” types for the reasons mentioned previously (poor categorization) and the fact that dividing these pollutions for sites with suitable and unstuitable substrate does not seem like a useful comparison. Why does it matter how pollution differs across suitable and unsuitable substrate? Substrate refers to whether there is sediment or rock, and is not impacted directly by “pollutions”. Further, comparing “amounts” of pollution in each category with pie charts isn’t appropriate when the levels of the different pollutions can’t be quantitatively compared among one another. I again suggest removing this figure, and field survey info, altogether.
382-383: Still not clear how “low risk” areas are defined or how an area can be at “risk of pollutants”.
385-386: Was the sighting of an actual meadow, or just seagrass wrack? See earlier comments regarding considering the presence of wrack as seagrass presence.
397-401: Again, your field surveys do not provide enough evidence to support these conclusions.
LInes 405-408: these lines are very repetitive of the abstract. Shorten/re-word?
418-419 and 420-421 related to field survey pollution data: You do not have results to back up these conclusions.
Review of Hastings et al. for Sustainability
Main comments
Overview:
This study developed a species distribution model (SDM) for predicting the distribution (or potential restoration area) for two seagrass species, Zostera marina and Z. noltii in the Republic of Ireland. Environmental and anthropogenic variables were used with the maximum entropy method to develop a suitability metric and maps for potential distribution. Physical variables were found to be more important than anthropogenic variables at the national scale (the scale of the SDM). The study also utilized 53 field surveys to assess a different suite of physical and anthropogenic variables and “pollutions” at a fine geographic scale.
This is an interesting study that presents a useful method for predicting seagrass extent or potential restoration areas on a broad scale when mapping is lacking. Seagrass distribution changes frequently and is difficult to map directly, so this is a useful tool. This study will be useful for the seagrass, blue carbon, and marine restoration communities, but several clarifications, and a few potentially large issues need to be addressed before the study can be published, and the writing could be made more concise and directly relevant to the study, especially in the introduction and discussion. The largest issue of the paper is surrounding the field surveys and the conclusions drawn from them, and I suggest this aspect of the study be removed entirely to improve the quality of the manuscript and the appropriateness of the conclusions drawn.
Title:
The title refers to ‘blue carbon’ but the study is about species distribution models for better describing seagrass distribution OR potential restoration areas. i.e. a second step is required after modeling to determine if seagrass is actually there, or why seagrass isn’t there (other limitations to seagrass survival and/or recruitment). The modeling itself does not result in maps of current distribution, just niche suitability, which is still useful. The title should be changed to better reflect what the study directly measures (which does not include carbon). How about "Modelling habitat suitability of seagrass ecosystems: a case study from the Republic of Ireland"?
Field surveys:
I’m not sure these field surveys should be included in the paper at all. They seem to weaken the overall quality of the study and don’t appear to add any useful findings. The interpretation and conclusions of the field surveys seems to go far beyond the actual data that was collected. The purpose of the 53 field surveys is unclear, mainly due to the definition of seagrass “presence” used in the study, which is incorrect. Seagrass present at the site that is “washed ashore” is not the same as observing live seagrass beds, but rather it should be described as “seagrass wrack”, which could be imported from distances near or far, and does not suggest living seagrass presence at the site. The study claims that 4 “new” seagrass sites were discovered, yet these are based on the presence of wrack, so should not be considered seagrass sites.
Methods surrounding the field surveys are vague, making assessment by the reader difficult. It is not clear how or why the sites were chosen, what methods were undertaken (e.g. transects? Quadrat measurements?), or whether the model output was compared to seagrass observations (Did you expect to find seagrass at the sites? Do findings agree with your modelled suitability distribution?).
The metrics recorded in the field survey are negatively biased and should be re-categorized to avoid this. For example, bioturbators should be called bioturbators, not “pests”. Macroalgal wrack should also be called what it is, as it is naturally deposited on coasts and beaches, and therefore should not be lumped in the “pollution” category, which suggests direct human impacts. I suggest re-categorizing (see specific comments).
Species Distribution Modelling:
-Caveats surrounding applying this model to determine current distribution and/or targeting sites for restoration are not critically discussed (see specific comments below). For example, it should be discussed that sites deemed “suitable” that do not currently have seagrass present should be critically evaluated to determine what is preventing natural seagrass recruitment and/or survival before restoration projects are given the green light to avoid restoration failure.
-An additional or substitute figure for the suitability maps would be useful to show areas that are and are not useful (based on a cut-off value of what is considered “suitable” – perhaps with levels – 0.9 and up is highly suitable, 0.75 and up is moderately suitable, and over 0.5 is likely suitable or something like that.
-More explanation on the meaning and interpretation of “percent contribution” and “permutation importance” are needed to help the reader understand why both metrics are presented, yet permutation importance is used as the final indicator of which variables are deemed more important for the SDM.
-Not enough info on methods and details of the species extent and physical variables used in the model. See specific comments, but for example, how were species distributions mapped previously and what year are those maps from? What is the resolution of bathymetry and other environmental data? What exactly are eco risk and eco status?
-No ground-truthing was used to test whether model predicted locations of new seagrass beds (though this is the claim – that the distribution is actually 4x that of current mapped extent). This is ok, but should be clearly stated.
Specific comments:
Abstract
Line 19: word choice “…or to identify suitable sites…” rather than “inform”
Line 20: briefly state why anthropogenic variables should be included (why is this an important aspect of your study that others have overlooked?).
Line 23: most readers won’t know what MaxEnt is. Re-phrase “Using the Maximum Entropy method for developing the SDM, …”
Line 26: Do you really mean “high standard deviation of SST”? A high variability in SST was important?
Line 30: It seems odd to put natural accumulation of wrack under the category of "pollution" with human-related variables of boat traffic and litter.
Line 30: "appeared to limit the distribution in county Cork." - Because you found few to no living seagrass sites in your survey, this statement is a speculation. It is not possible to say that these are limiting the distribution when you didn't survey sites with healthy Zostera meadows (which could have also had these stressors, but survived).
Line 31: Were seagrass beds expected at these survey sites based on the SDM?
Line 32-33: Were results of the SDM ground-truthed at all? Also, be cautious in wording about “suitable” areas with no seagrass being used for restoration (there could be reasons why seagrass isn’t already there).
Introduction
Line 47: “are” should be replaced by “is”
Line 49: “satisfy” should be “satisfies”
Line 52: “implementing” should be “incorporating”
Line 55-58: sentence tenses odd and “bioindicators…” phrasing doesn’t fit grammatically
Line 48-67: Can these two paragraphs be combined in shortened? Each paragraph takes a separate approach to justifying the need for better mapping, but seem disjointed when presented separately.
Line 68-69: for studying “all” taxa? This seems extremely broad. For studying a “variety” of taxa?
Lines 76-77: Gradients and differences of what? Bathymetry? Salinity?
Line 83: “they are not the sole driver” – isn’t this what you are testing? Rephrase to “they may not be the sole driver”
Lines 84-96: Sentences in this paragraph seem very repetitive. Can you go more into the mechanisms and examples of how anthropogenic drivers may limit seagrass distribution, rather than restating that anthropogenic variables aren’t usually included in SDMs? The latter can be stated one time with multiple references.
Lines 107-109: I don’t think questions 3 and 4 are answered in this study. I would remove these.
Methods:
Lines 125-129: What does “verified” mean in this context? Please explain. How were the seagrass distribution areas mapped/estimated? What year are they from? Were they measured via satellite and ground trothed? Mapped manually?
Figure 1: Geographic labels on your maps would be helpful (“Republic of Ireland” and “county Cork”). A box around Cork in map “a” would also be helpful.
Lines 140-141: It seems odd to convert polygons to points, considering points cover only a small area compared to the potentially large areas covered by polygons. Was seagrass area lost due to conversion of polygons to points? Or were smaller polygons created to overlap the spatial coverage of larger polygons to ensure points covered the majority of the seagrass extent? Please clarify how you were able to maintain true area when converting polygons to points.
Line 146: Much more detail is needed about the field surveys. Please briefly describe this “predominant method” used in Ireland for readers who are unfamiliar. How much area was covered at each site? How were sites chosen? Were these sites compared to the model predictions later on? Were transects used? Quadrats? Did you measure density? How do you define “presence” of seagrass? Much more detail is needed here.
Lines 147-148: sentence structure - "Following the recommendations of the EPA, surveys took place..."
Line 151: Why was 1 km-resolution chosen? Did this match many of the original spatial data layers?
Line 156-157: It bioturbators are generally not considered “pests”, as this gives a negative (biased) connotation to the fauna naturally found in the sediment. Similarly, “seaweed piling” is usually naturally accumulated macroalgal wrack, not a result of human “pollution”. More appropriate categories would be substrate, exposure, pollution, macroalgal wrack, and infauna.
Line 159-160: “exposure and substrate” – details? How were exposure and substrate categorized and determined in the field?
Line 170: “ascertain” should be changed to “predict” or “estimate” since you did not confirm your model predictions of distribution.
Table 1:
-Each statement in the “Justification” column should have references associated with it.
-Methodology column – could this column be better titled as “Spatial Data Pre-Processing” or something similar?
-More information is needed for many of the layers either here or in the text - e.g. what was the original resolution of bathymetry and other data layers? (Briefly) What data collection methods were used to create the layers? What year were the spatial data generated? This info is key for the reader to understand how the environmental variables relate to spatial and temporal scales as compared to the species distribution data and suitability scores generated based on 1 km resolution.
-“WFD” – what does this stand for?
-“Eco status” and “risk status” – it is not clear from the description what these mean or what data were used to generate them.
- "Exposure" usually refers to wave exposure, but it sounds like this is more of a categorical assessment based on anthropogenic risk? Please clarify what this is, as well as Eco Status and Risk Status. How can you get a physical predictor by combining two anthropogenic predictors?
Line 183 – reference?
Lines 184-185: This describes your "suitability" indicator, correct? What is your cut-off for what is considered "suitable" habitat? Is it >0.5? Higher? This should be clearly stated, then applied to specify estimated distribution area in a figure in addition to figure 2 to show maps of "suitable" area (rather than only a range of suitability from 0-1, which is what is shown now). It is difficult to tell from the color gradient how much of figure 2 area is actually “suitable” (e.g. over 0.5). A new set of maps (similar in extent to Figure 2) showing binned levels of suitability would be very useful for showing where suitable areas are, and how they scored compared to one another. For example, a “very high suitability” category for >0.9, a “high suitability” category for 0.75-0.89, and a “moderate suitability” category for >0.5-0.74 (each symbolized in the maps) to show areas of highest priority in Ireland for further investigation (i.e. for searching for unmapped beds, or assessing the sites’ potential for restoration).
Lines 189-192: Sentence long and a bit confusing. Please reword/clarify.
Line 188: Why use all three methods for evaluation? When they give differing results (table 2), which value is more important? Can you briefly describe how to assess these values compared to one another to ease the reader's interpretation of the data (e.g. Table 2). In the results, it seems percent contribution is essentially entirely ignored and permutation importance is used as the default indicator to determine the ranking of important drivers. Why?
Line 203: Again, why was 1km resolution used? Why not finer? Can you provide a justification for this earlier in the methods?
Results
Line 218: Why focus on this instead of percent contribution? Dredge fishing and eco status each have ~10% contribution – is this not important if permutation importance percentages don't reflect these? Please explain how each of these parameters (percent contr vs permutation importance) is assessed, especially when they contradict one another.
Line 223: If there is a “positive” response to bathymetry between 3-10 m, wouldn’t that mean either 3 or 10 would be the ideal bathymetry, not 5 m? Please clarify.
Line 225: “is” should be changed to “was”
Line 228: This seems like an odd/unexpected result. It is counterintuitive to see that a smaller variability in temperature would be negatively associated with seagrass distribution. You do not directly address this result in the discussion. Shouldn’t areas with fewer temperature fluctuations (i.e. less extremes of hot and cold temperature stress) be more suitable for seagrass? Can you speculate or find literature to help explain this result in the discussion (I also commented on below in the discussion).
Line 233: “is” should be changed to “was”
Figure 2
- These maps show a range of suitability - which areas are considered "suitable" vs not? Though it is useful to see the gradient of 1 to 0 for the color symbology, as-is it cannot be utilized to identify potential distribution. An additional figure showing suitability (for the same panel areas for each species) would be useful, with estimates of areal extent of potential distribution for each.
- Can you add labels to the inset boxes and matching zoomed in panels (e.g. "A" on the inside of the northernmost box, with a matching "A" in the top left zoomed panel), also "Ireland" to main map.
-The resolution looks a little blurry for the color and text - can this be improved (i.e. is it related to figure processing) or is this related to the resolution of the modelling?
- The caption should allow the figure to stand alone from the main text. Please expand on this caption to briefly describe how the figure was generated, e.g. "Habitat suitability... based on physical and anthropogenic variables assessed through MaxEnt"
Section 3.2 Field surveys
- Seagrass that is detached or "washed ashore" should be referred to as “seagrass wrack” and should not be confused with living seagrass beds. It does not indicate presence of growing seagrass, and without further investigation it is very difficult, if not impossible, to determine where the seagrass washed in from. Seagrass wrack should not ever be used as a proxy for seagrass “presence”.
-What is a "sample" in this context? Is it a ramet of detached seagrass? This is simply wrack, and does not give a proxy of proximity. Seagrass wrack can be transported over very long distances in a short amount of time. It is only appropriate to report seagrass presence where seagrass was actually seen growing, anchored and alive in the sediments. If no seagrass was found in the surveys (which is how it sounds), the aspects of the study related to the field surveys should not be included, because sites that represent both seagrass presence and absence were not sampled, and therefore links between measured variables and seagrass suitability cannot be made. This section should either be entirely removed (along with methods and discussion related to field surveys) or be re-written to clearly reflect that these surveys are qualitatively informative only (and information on “samples” is not useful). The surveys could be briefly mentioned in that they can be used to suggest potential relationships between the variables observed (e.g. macroalgal wrack presence and lack of living seagrass) that should be investigated in the future, but cannot be used to make conclusions from this study (and this should be clearly and explicitly reflected in the discussion and abstract). This would be a much stronger paper without including the field surveys and I highly suggest you remove this content.
Table 2: Again, why is permutation importance discussed as the metric that determines driver importance, while percent contribution is largely ignored?
Figures 3 and 4. Move to supplementary material. As things are, there is not much in the results about the response curves, and no information on how they are generated or what they are based on. This information should be added in the text, and I think both response curve figures are more appropriate in the supplementary materials if they will not be discussed in more detail. You mention the 3-10m optimal bathymetry range for Z. marina, but it is very hard to distinguish that from this figure because individual panels are quite small.
Line 263: Very vague first sentence. Can this be made more informative?
Lines 263-267: Again, (if not completely removed) this section should be altered based on comments in the methods. Seaweed wrack is not pollution. Bioturbators are not "pests".
Line 266 – “half deemed at risk” – what does this mean? What is it based on?
Lines 269-270: Avoid referring to figures in this way. "It can be seen from Figure 6a" - instead just state "Seaweed pilings were the...(Figure 6a)."
Figure 5: very misleading if seagrass “presence” referred to seagrass wrack observed at the site…
Discussion
General:
It is important to discuss the limitations of your study and your methods -
-Why didn't you "ground truth" any of the predictions from your model to see if seagrass was present where your suitability scores were high?
-In the intro, you discuss the applicability of the SDM for identifying current distribution/range, or identifying areas fit for restoration if the current range doesn't actually cover those suitable areas. Isn't this risky? If seagrass isn't present in areas deemed "suitable" by the model, perhaps there are other factors limiting seagrass distribution that should be assessed to ensure restoration success. Restoration success may be limited by factors not measured in this study, not mapped at relevant scales, or for which there is no data. Survival and recruitment may be limited by seed availability (proximity and/or connectivity to other meadows) or other factors. These should be clearly addressed throughout the paper as a caution. Areas deemed "suitable" by your model may in reality be risky sites for restoration without further investigation. It begs the question why isn't seagrass there already?
-Though you identified a range of suitability levels, some "suitable" areas may not actually have seagrass. What else could be limiting seagrass establishment or survival in these areas? You mention macroalgal wrack, but there are many potential factors that you didn't model, or as you touch on can't be modelled at broad scales, that could impact recruitment and survival. Before undertaking large restoration projects in "suitable" areas, what other factors would be assessed or tested in addition to macroalgal wrack presence to ensure success?
-Some of the discussion text that is very broad compared to the study focused could be cut to discuss in better depth the limitations listed above.
Line 279: Isn't your aim to determine which are most important, not just which type? I recommend rewording: "...identify key physical and anthropogenic environmental drivers..."
Line 284: lots of jargon. Should this be explained in the methods?
Line 285: Are these values reported in the results? If not, add them.
287-292: It seems a lot of this should be moved to the methods and results before discussing it here.
299-324: It seems like the focus of these paragraphs is other research, with mentions of your results as sort of a side note. Can these paragraphs be changed to better highlight the importance of your research in these contexts?
344-349: This seems a bit off topic (beyond the scope of this study), but perhaps could be mentioned just briefly how SST relates to growth and blue carbon (with appropriate references).
Line 325: “it is clear…” – strong wording, please rephrase.
Lines 325-340: I recommend deleting this paragraph. Unattached seagrass (i.e wrack) does not indicate presence nearby. It being “unlikely” to have traveled (lines 329-330) is purely speculation, and not supported by data on seagrass wrack, which can drift great distances. Large masses of seagrass wrack can accumulate far from meadows due to currents and waves, water circulation direction, fetch, and so on.
Line 357: remove “successfully”
Line 359: This seems very odd. Can you explain or speculate on why? Is there anything in the literature to support that more variability in temperature is somehow beneficial for seagrass growth/survival? I would expect the opposite – extreme highs and lows would be associated with intolerance due to heat and cold stress.
Figure 6: Again, I think the study would be better without the inclusion of the field surveys. Pie charts don’t seem appropriate for comparing “pollution” types for the reasons mentioned previously (poor categorization) and the fact that dividing these pollutions for sites with suitable and unstuitable substrate does not seem like a useful comparison. Why does it matter how pollution differs across suitable and unsuitable substrate? Substrate refers to whether there is sediment or rock, and is not impacted directly by “pollutions”. Further, comparing “amounts” of pollution in each category with pie charts isn’t appropriate when the levels of the different pollutions can’t be quantitatively compared among one another. I again suggest removing this figure, and field survey info, altogether.
382-383: Still not clear how “low risk” areas are defined or how an area can be at “risk of pollutants”.
385-386: Was the sighting of an actual meadow, or just seagrass wrack? See earlier comments regarding considering the presence of wrack as seagrass presence.
397-401: Again, your field surveys do not provide enough evidence to support these conclusions.
LInes 405-408: these lines are very repetitive of the abstract. Shorten/re-word?
418-419 and 420-421 related to field survey pollution data: You do not have results to back up these conclusions.
Author Response
*** Denotes author responses
Overview:
This study developed a species distribution model (SDM) for predicting the distribution (or potential restoration area) for two seagrass species, Zostera marina and Z. noltii in the Republic of Ireland. Environmental and anthropogenic variables were used with the maximum entropy method to develop a suitability metric and maps for potential distribution. Physical variables were found to be more important than anthropogenic variables at the national scale (the scale of the SDM). The study also utilized 53 field surveys to assess a different suite of physical and anthropogenic variables and “pollutions” at a fine geographic scale.
This is an interesting study that presents a useful method for predicting seagrass extent or potential restoration areas on a broad scale when mapping is lacking. Seagrass distribution changes frequently and is difficult to map directly, so this is a useful tool. This study will be useful for the seagrass, blue carbon, and marine restoration communities, but several clarifications, and a few potentially large issues need to be addressed before the study can be published, and the writing could be made more concise and directly relevant to the study, especially in the introduction and discussion. The largest issue of the paper is surrounding the field surveys and the conclusions drawn from them, and I suggest this aspect of the study be removed entirely to improve the quality of the manuscript and the appropriateness of the conclusions drawn.
Title:
The title refers to ‘blue carbon’ but the study is about species distribution models for better describing seagrass distribution OR potential restoration areas. i.e. a second step is required after modeling to determine if seagrass is actually there, or why seagrass isn’t there (other limitations to seagrass survival and/or recruitment). The modeling itself does not result in maps of current distribution, just niche suitability, which is still useful. The title should be changed to better reflect what the study directly measures (which does not include carbon). How about "Modelling habitat suitability of seagrass ecosystems: a case study from the Republic of Ireland"?
***A good point. We have tempered the implications of the title, but have opted to include the novel aspects of the model. It now reads “Assessing the impact of physical and anthropogenic environmental factors in determining the habitat suitability of seagrass ecosystems”
Field surveys:
I’m not sure these field surveys should be included in the paper at all. They seem to weaken the overall quality of the study and don’t appear to add any useful findings. The interpretation and conclusions of the field surveys seems to go far beyond the actual data that was collected. The purpose of the 53 field surveys is unclear, mainly due to the definition of seagrass “presence” used in the study, which is incorrect. Seagrass present at the site that is “washed ashore” is not the same as observing live seagrass beds, but rather it should be described as “seagrass wrack”, which could be imported from distances near or far, and does not suggest living seagrass presence at the site. The study claims that 4 “new” seagrass sites were discovered, yet these are based on the presence of wrack, so should not be considered seagrass sites.
Methods surrounding the field surveys are vague, making assessment by the reader difficult. It is not clear how or why the sites were chosen, what methods were undertaken (e.g. transects? Quadrat measurements?), or whether the model output was compared to seagrass observations (Did you expect to find seagrass at the sites? Do findings agree with your modelled suitability distribution?).
The metrics recorded in the field survey are negatively biased and should be re-categorized to avoid this. For example, bioturbators should be called bioturbators, not “pests”. Macroalgal wrack should also be called what it is, as it is naturally deposited on coasts and beaches, and therefore should not be lumped in the “pollution” category, which suggests direct human impacts. I suggest re-categorizing (see specific comments).
***We’ve taken these constructive comments on board, and upon reflection we agree with you. We felt given the significant workload involved in completing these, we wanted to include them; however, we acknowledge the qualitative consideration of the results. Unfortunately, we could not undertake in situ validation of Z. marina due to adverse conditions for snorkelling. We have also explored the possibility of using high resolution satellite imagery to validate this; however, the patterns observed in other studies were not clear, and we could not distinguish between seagrass and other macroalgae, which would introduce more uncertainties into any conclusions in our opinion. Therefore, we decided to omit this. We still make passing comment to this in the discussion in a qualitative manner, but this no longer constitutes a significant part of the work.
Species Distribution Modelling:
-Caveats surrounding applying this model to determine current distribution and/or targeting sites for restoration are not critically discussed (see specific comments below). For example, it should be discussed that sites deemed “suitable” that do not currently have seagrass present should be critically evaluated to determine what is preventing natural seagrass recruitment and/or survival before restoration projects are given the green light to avoid restoration failure.
***A good point. We have now elaborated on this in the discussion to really emphasise that suitable does not necessarily mean it will be successful.
-An additional or substitute figure for the suitability maps would be useful to show areas that are and are not useful (based on a cut-off value of what is considered “suitable” – perhaps with levels – 0.9 and up is highly suitable, 0.75 and up is moderately suitable, and over 0.5 is likely suitable or something like that.
***This is a good idea, but upon implementation the maps were broadly similar, as high and very high suitability were clearly distinguishable. To ensure that this figure is easier to read, we have ensured they are generated at a suitable dpi, and they now appear much cleaner.
-More explanation on the meaning and interpretation of “percent contribution” and “permutation importance” are needed to help the reader understand why both metrics are presented, yet permutation importance is used as the final indicator of which variables are deemed more important for the SDM.
***We have now rewritten this section and clarified this. Here we identify the limitations of using percent contribution, which leads to our preferred use of permutation importance. However, many SDM studies still only report percent contribution, so we felt it best to report both statistics to rigor.
-Not enough info on methods and details of the species extent and physical variables used in the model. See specific comments, but for example, how were species distributions mapped previously and what year are those maps from? What is the resolution of bathymetry and other environmental data? What exactly are eco risk and eco status?
***We have now provided more information on the response data and environmental variables in Table 1
-No ground-truthing was used to test whether model predicted locations of new seagrass beds (though this is the claim – that the distribution is actually 4x that of current mapped extent). This is ok, but should be clearly stated.
***We do mention this in the discussion, but we have now tempered this statement and have clarified this throughout the manuscript.
Specific comments:
Abstract
Line 19: word choice “…or to identify suitable sites…” rather than “inform”
***Changed
Line 20: briefly state why anthropogenic variables should be included (why is this an important aspect of your study that others have overlooked?).
***We have now added an explanation.
Line 23: most readers won’t know what MaxEnt is. Re-phrase “Using the Maximum Entropy method for developing the SDM, …”
***We have now added this extra information
Line 26: Do you really mean “high standard deviation of SST”? A high variability in SST was important?
***Yes, this was a result we identified, and is clear from the response curves (a reason we have decided to retain these curves in the submission). We have also speculated as to why this might be in the discussion.
Line 30: It seems odd to put natural accumulation of wrack under the category of "pollution" with human-related variables of boat traffic and litter.
***We have now removed this in line with your other comments
Line 30: "appeared to limit the distribution in county Cork." - Because you found few to no living seagrass sites in your survey, this statement is a speculation. It is not possible to say that these are limiting the distribution when you didn't survey sites with healthy Zostera meadows (which could have also had these stressors, but survived).
***This has now been removed following the rewrite
Line 31: Were seagrass beds expected at these survey sites based on the SDM?
***This has now been removed during the rewrite
Line 32-33: Were results of the SDM ground-truthed at all? Also, be cautious in wording about “suitable” areas with no seagrass being used for restoration (there could be reasons why seagrass isn’t already there).
***We have clarified this in the document, and elaborate on why we could not ‘fully’ ground truth in the discussion
Introduction
Line 47: “are” should be replaced by “is”
***Changed
Line 49: “satisfy” should be “satisfies”
***Changed
Line 52: “implementing” should be “incorporating”
***Changed
Line 55-58: sentence tenses odd and “bioindicators…” phrasing doesn’t fit grammatically
***We have now split this sentence and changed some of the terminology on the water quality example
Line 48-67: Can these two paragraphs be combined in shortened? Each paragraph takes a separate approach to justifying the need for better mapping, but seem disjointed when presented separately.
***A good point. We have shortened these two paragraphs and have now merged the paragraphs to provide one focal justification for the need for better mapping
Line 68-69: for studying “all” taxa? This seems extremely broad. For studying a “variety” of taxa?
***Point taken. In fairness, we would argue this method has been used for all taxa, including bacteria and viruses, but given this isn’t the aim of the study we have changed this.
Lines 76-77: Gradients and differences of what? Bathymetry? Salinity?
***This has now been clarified
Line 83: “they are not the sole driver” – isn’t this what you are testing? Rephrase to “they may not be the sole driver”
***True. There is certainly a lot of literature that suggests this is the case, but not in an SDM context. Therefore, we have changed this to may not be
Lines 84-96: Sentences in this paragraph seem very repetitive. Can you go more into the mechanisms and examples of how anthropogenic drivers may limit seagrass distribution, rather than restating that anthropogenic variables aren’t usually included in SDMs? The latter can be stated one time with multiple references.
***A good point. We have now reworded this paragraph, and have provided an example on eutrophication
Lines 107-109: I don’t think questions 3 and 4 are answered in this study. I would remove these.
***Again, on reflection we would agree with you. We have removed these questions from the article to focus on the first two.
Methods:
Lines 125-129: What does “verified” mean in this context? Please explain. How were the seagrass distribution areas mapped/estimated? What year are they from? Were they measured via satellite and ground trothed? Mapped manually?
***We have now provided information from these studies
Figure 1: Geographic labels on your maps would be helpful (“Republic of Ireland” and “county Cork”). A box around Cork in map “a” would also be helpful.
***We have now improved this figure, and removed Cork as it no longer features as a part of the analysis
Lines 140-141: It seems odd to convert polygons to points, considering points cover only a small area compared to the potentially large areas covered by polygons. Was seagrass area lost due to conversion of polygons to points? Or were smaller polygons created to overlap the spatial coverage of larger polygons to ensure points covered the majority of the seagrass extent? Please clarify how you were able to maintain true area when converting polygons to points.
***This is unfortunately a trade-off of the SDM framework that requires point data as an input for the response variable. As we state in the manuscript, we did look into other methods, including overlaying a fishnet grid over the polygons; however this overestimated the distribution much more than converting the polygons to points underestimated given the 1km resolution of the environmental data we were coupling the points with. All verified seagrass meadows were smaller than 1km2, and so we were confident that we were not drastically reducing the true area of these meadows, as the centroid often fell within the grid where the majority of the seagrass was contained. However, we are confident having explored various options of data transformation we have chosen the most reliable method, and it is beyond the scope of this research to provide a comparison.
Line 146: Much more detail is needed about the field surveys. Please briefly describe this “predominant method” used in Ireland for readers who are unfamiliar. How much area was covered at each site? How were sites chosen? Were these sites compared to the model predictions later on? Were transects used? Quadrats? Did you measure density? How do you define “presence” of seagrass? Much more detail is needed here.
***We have now removed this section from the manuscript
Lines 147-148: sentence structure - "Following the recommendations of the EPA, surveys took place..."
***This has now been removed from the manuscript
Line 151: Why was 1 km-resolution chosen? Did this match many of the original spatial data layers?
***Yes, this has been clarified
Line 156-157: It bioturbators are generally not considered “pests”, as this gives a negative (biased) connotation to the fauna naturally found in the sediment. Similarly, “seaweed piling” is usually naturally accumulated macroalgal wrack, not a result of human “pollution”. More appropriate categories would be substrate, exposure, pollution, macroalgal wrack, and infauna.
***We have now removed this part of the analysis following your recommendation, so we have removed this text. When we discuss bioturbators in the article, we ensure not to refer to them as pests, and we have clarified any text on algal wrack.
Line 159-160: “exposure and substrate” – details? How were exposure and substrate categorized and determined in the field?
***Same as above comment
Line 170: “ascertain” should be changed to “predict” or “estimate” since you did not confirm your model predictions of distribution.
***Changed
Table 1:
-Each statement in the “Justification” column should have references associated with it.
***We have now added increased justification and explanation, and citations where necessarily.
-Methodology column – could this column be better titled as “Spatial Data Pre-Processing” or something similar?
***This has now been changed
-More information is needed for many of the layers either here or in the text - e.g. what was the original resolution of bathymetry and other data layers? (Briefly) What data collection methods were used to create the layers? What year were the spatial data generated? This info is key for the reader to understand how the environmental variables relate to spatial and temporal scales as compared to the species distribution data and suitability scores generated based on 1 km resolution.
***We have now added this information to the table. Not all information on years was available, and polygon data does not require a spatial reference.
-“WFD” – what does this stand for?
***This has now been explained in the table caption.
-“Eco status” and “risk status” – it is not clear from the description what these mean or what data were used to generate them.
***We have now provided more information in the table on these
- "Exposure" usually refers to wave exposure, but it sounds like this is more of a categorical assessment based on anthropogenic risk? Please clarify what this is, as well as Eco Status and Risk Status. How can you get a physical predictor by combining two anthropogenic predictors?
***A good point, and we have now clarified what we mean by these variables so that it is clear exposure relates to wave exposure.
Line 183 – reference?
***Reference added
Lines 184-185: This describes your "suitability" indicator, correct? What is your cut-off for what is considered "suitable" habitat? Is it >0.5? Higher? This should be clearly stated, then applied to specify estimated distribution area in a figure in addition to figure 2 to show maps of "suitable" area (rather than only a range of suitability from 0-1, which is what is shown now). It is difficult to tell from the color gradient how much of figure 2 area is actually “suitable” (e.g. over 0.5). A new set of maps (similar in extent to Figure 2) showing binned levels of suitability would be very useful for showing where suitable areas are, and how they scored compared to one another. For example, a “very high suitability” category for >0.9, a “high suitability” category for 0.75-0.89, and a “moderate suitability” category for >0.5-0.74 (each symbolized in the maps) to show areas of highest priority in Ireland for further investigation (i.e. for searching for unmapped beds, or assessing the sites’ potential for restoration).
***This is a good idea, but upon implementation the maps were broadly similar, as high and very high suitability were clearly distinguishable. To ensure that this figure is easier to read, we have ensured they are generated at a suitable dpi, and they now appear much cleaner.
Lines 189-192: Sentence long and a bit confusing. Please reword/clarify.
***This has been clarified and addressed earlier with your major comments
Line 188: Why use all three methods for evaluation? When they give differing results (table 2), which value is more important? Can you briefly describe how to assess these values compared to one another to ease the reader's interpretation of the data (e.g. Table 2). In the results, it seems percent contribution is essentially entirely ignored and permutation importance is used as the default indicator to determine the ranking of important drivers. Why?
***This has been clarified and addressed earlier with your major comments
Line 203: Again, why was 1km resolution used? Why not finer? Can you provide a justification for this earlier in the methods?
***This has been addressed in the discussion. This was the finest resolution we could use the available data at, without oversimplifying some of the coarser data.
Results
Line 218: Why focus on this instead of percent contribution? Dredge fishing and eco status each have ~10% contribution – is this not important if permutation importance percentages don't reflect these? Please explain how each of these parameters (percent contr vs permutation importance) is assessed, especially when they contradict one another.
***We have now addressed this more specifically in the methods. We are loathed to remove percent contribution, as it is still regularly used to assess environmental importance, but it is not considered best practice. In fact, the large differences highlight the importance of using both metrics, particularly as permutation importance is considered more robust than percent contribution.
Line 223: If there is a “positive” response to bathymetry between 3-10 m, wouldn’t that mean either 3 or 10 would be the ideal bathymetry, not 5 m? Please clarify.
*** We have now clarified the wording of this so it’s not confusing. We appreciate the manner in which we have written this appears to assume a linear relationship, which is not the case
Line 225: “is” should be changed to “was”
***Changed
Line 228: This seems like an odd/unexpected result. It is counterintuitive to see that a smaller variability in temperature would be negatively associated with seagrass distribution. You do not directly address this result in the discussion. Shouldn’t areas with fewer temperature fluctuations (i.e. less extremes of hot and cold temperature stress) be more suitable for seagrass? Can you speculate or find literature to help explain this result in the discussion (I also commented on below in the discussion).
***This assumes that an increased range takes values beyond the physiological tolerances of the range. The temperature range is quite broad -1 to 25, meaning a higher standard deviation could provide more suitable conditions to locations that were not previously suitable. We have speculated this in the discussion but suggest future research to investigate why this is the case.
Line 233: “is” should be changed to “was”
***Changed
Figure 2
- These maps show a range of suitability - which areas are considered "suitable" vs not? Though it is useful to see the gradient of 1 to 0 for the color symbology, as-is it cannot be utilized to identify potential distribution. An additional figure showing suitability (for the same panel areas for each species) would be useful, with estimates of areal extent of potential distribution for each.
***This is a good idea, but upon implementation the maps were broadly similar, as high and very high suitability were clearly distinguishable. To ensure that this figure is easier to read, we have ensured they are generated at a suitable dpi, and they now appear much cleaner.
- Can you add labels to the inset boxes and matching zoomed in panels (e.g. "A" on the inside of the northernmost box, with a matching "A" in the top left zoomed panel), also "Ireland" to main map.
***We have now provided context for the boxes
-The resolution looks a little blurry for the color and text - can this be improved (i.e. is it related to figure processing) or is this related to the resolution of the modelling?
***This is in part due to the resolution of the modelling and the broad extent. We have ensured the dpi is now up to publication quality.
- The caption should allow the figure to stand alone from the main text. Please expand on this caption to briefly describe how the figure was generated, e.g. "Habitat suitability... based on physical and anthropogenic variables assessed through MaxEnt"
***We have now expanded the figure caption to include this information
Section 3.2 Field surveys
- Seagrass that is detached or "washed ashore" should be referred to as “seagrass wrack” and should not be confused with living seagrass beds. It does not indicate presence of growing seagrass, and without further investigation it is very difficult, if not impossible, to determine where the seagrass washed in from. Seagrass wrack should not ever be used as a proxy for seagrass “presence”.
***We have now removed this section from the analysis. Given our knowledge of the area, we would be relatively comfortable stating that these are from nearby sources, but we appreciate we cannot make these statements without further evidence. We did explore the use of high resolution remote sensing to validate Z.marina, but we did not observe any distinguishing patterns that would allow us to quantify this.
-What is a "sample" in this context? Is it a ramet of detached seagrass? This is simply wrack, and does not give a proxy of proximity. Seagrass wrack can be transported over very long distances in a short amount of time. It is only appropriate to report seagrass presence where seagrass was actually seen growing, anchored and alive in the sediments. If no seagrass was found in the surveys (which is how it sounds), the aspects of the study related to the field surveys should not be included, because sites that represent both seagrass presence and absence were not sampled, and therefore links between measured variables and seagrass suitability cannot be made. This section should either be entirely removed (along with methods and discussion related to field surveys) or be re-written to clearly reflect that these surveys are qualitatively informative only (and information on “samples” is not useful). The surveys could be briefly mentioned in that they can be used to suggest potential relationships between the variables observed (e.g. macroalgal wrack presence and lack of living seagrass) that should be investigated in the future, but cannot be used to make conclusions from this study (and this should be clearly and explicitly reflected in the discussion and abstract). This would be a much stronger paper without including the field surveys and I highly suggest you remove this content.
***On reflection, we agree and have removed this section completely.
Table 2: Again, why is permutation importance discussed as the metric that determines driver importance, while percent contribution is largely ignored?
***We have addressed this comment earlier
Figures 3 and 4. Move to supplementary material. As things are, there is not much in the results about the response curves, and no information on how they are generated or what they are based on. This information should be added in the text, and I think both response curve figures are more appropriate in the supplementary materials if they will not be discussed in more detail. You mention the 3-10m optimal bathymetry range for Z. marina, but it is very hard to distinguish that from this figure because individual panels are quite small.
***We have decided that given the restructure, it was important to keep the response curves. We have subsequently added more discussion on this in the results, and attempted to increase the readability of the panels, although given the range of values for some variables, unfortunately this was difficult to implement and we don’t have access to software that would allow to ‘zoom in’ on the panels to improve this
Line 263: Very vague first sentence. Can this be made more informative?
***This section has been removed
Lines 263-267: Again, (if not completely removed) this section should be altered based on comments in the methods. Seaweed wrack is not pollution. Bioturbators are not "pests".
***This section has been removed
Line 266 – “half deemed at risk” – what does this mean? What is it based on?
***This section has been removed
Lines 269-270: Avoid referring to figures in this way. "It can be seen from Figure 6a" - instead just state "Seaweed pilings were the...(Figure 6a)."
***This section has been removed
Figure 5: very misleading if seagrass “presence” referred to seagrass wrack observed at the site…
***Figure has been removed
Discussion
General:
It is important to discuss the limitations of your study and your methods -
***We have undertaken quite a big restructure of the discussion, specifically addressing the limitations.
-Why didn't you "ground truth" any of the predictions from your model to see if seagrass was present where your suitability scores were high?
***That was the rationale behind the field surveys, but felt we couldn’t use it as ground truth without the marina observations. We now specifically include a paragraph on this in the discussion
-In the intro, you discuss the applicability of the SDM for identifying current distribution/range, or identifying areas fit for restoration if the current range doesn't actually cover those suitable areas. Isn't this risky? If seagrass isn't present in areas deemed "suitable" by the model, perhaps there are other factors limiting seagrass distribution that should be assessed to ensure restoration success. Restoration success may be limited by factors not measured in this study, not mapped at relevant scales, or for which there is no data. Survival and recruitment may be limited by seed availability (proximity and/or connectivity to other meadows) or other factors. These should be clearly addressed throughout the paper as a caution. Areas deemed "suitable" by your model may in reality be risky sites for restoration without further investigation. It begs the question why isn't seagrass there already?
***In fairness, our accuracy metrics using our 10-fold cross validation reports high accuracy <0.9, suggesting we do successfully predict presence and pseudo-absence data. We had also made reference to this in the discussion, but we have clarified and reference this more in the discussion in a more succinct manner
-Though you identified a range of suitability levels, some "suitable" areas may not actually have seagrass. What else could be limiting seagrass establishment or survival in these areas? You mention macroalgal wrack, but there are many potential factors that you didn't model, or as you touch on can't be modelled at broad scales, that could impact recruitment and survival. Before undertaking large restoration projects in "suitable" areas, what other factors would be assessed or tested in addition to macroalgal wrack presence to ensure success?
***We have eluded to this in the final paragraph of the discussion, particularly in relation to our observations (with supplementary information to support this)
-Some of the discussion text that is very broad compared to the study focused could be cut to discuss in better depth the limitations listed above.
***With the rewrite this should be more succinct now.
Line 279: Isn't your aim to determine which are most important, not just which type? I recommend rewording: "...identify key physical and anthropogenic environmental drivers..."
***Changed
Line 284: lots of jargon. Should this be explained in the methods?
*** We rewritten this and moved to the results. We feel free that the statement is okay here and not needed in the methods, as if readers are unfamiliar with SDM, it could cause confusion with the methods as it’s such a short statement it may introduce uncertainty about how we’ll delineating suitable and unsuitable.
Line 285: Are these values reported in the results? If not, add them.
***This has now been moved to the results
287-292: It seems a lot of this should be moved to the methods and results before discussing it here.
*** We rewritten this and moved to the results. We feel free that the statement is okay here and not needed in the methods, as if readers are unfamiliar with SDM, it could cause confusion with the methods as it’s such a short statement it may introduce uncertainty about how we’ll delineating suitable and unsuitable.
299-324: It seems like the focus of these paragraphs is other research, with mentions of your results as sort of a side note. Can these paragraphs be changed to better highlight the importance of your research in these contexts?
***We have substantially rewritten this paragraph to address these point.
344-349: This seems a bit off topic (beyond the scope of this study), but perhaps could be mentioned just briefly how SST relates to growth and blue carbon (with appropriate references).
***We have rewritten this section, and in doing so feel we address this more appropriately.
Line 325: “it is clear…” – strong wording, please rephrase.
***This has been removed
Lines 325-340: I recommend deleting this paragraph. Unattached seagrass (i.e wrack) does not indicate presence nearby. It being “unlikely” to have traveled (lines 329-330) is purely speculation, and not supported by data on seagrass wrack, which can drift great distances. Large masses of seagrass wrack can accumulate far from meadows due to currents and waves, water circulation direction, fetch, and so on.
***This has been removed.
Line 357: remove “successfully”
***This has been removed.
Line 359: This seems very odd. Can you explain or speculate on why? Is there anything in the literature to support that more variability in temperature is somehow beneficial for seagrass growth/survival? I would expect the opposite – extreme highs and lows would be associated with intolerance due to heat and cold stress.
***We have added a speculation and suggested future research is warranted. Comment addressed in more detail earlier in these responses.
Figure 6: Again, I think the study would be better without the inclusion of the field surveys. Pie charts don’t seem appropriate for comparing “pollution” types for the reasons mentioned previously (poor categorization) and the fact that dividing these pollutions for sites with suitable and unstuitable substrate does not seem like a useful comparison. Why does it matter how pollution differs across suitable and unsuitable substrate? Substrate refers to whether there is sediment or rock, and is not impacted directly by “pollutions”. Further, comparing “amounts” of pollution in each category with pie charts isn’t appropriate when the levels of the different pollutions can’t be quantitatively compared among one another. I again suggest removing this figure, and field survey info, altogether.
***Upon reflection and your comments, we have removed this section
382-383: Still not clear how “low risk” areas are defined or how an area can be at “risk of pollutants”.
***This should be clarified given the rewrite
385-386: Was the sighting of an actual meadow, or just seagrass wrack? See earlier comments regarding considering the presence of wrack as seagrass presence.
***This has been removed.
397-401: Again, your field surveys do not provide enough evidence to support these conclusions.
***This has been removed.
LInes 405-408: these lines are very repetitive of the abstract. Shorten/re-word?
***Given these are in the conclusion, we feel it is fine to keep this text as it is, as it succinctly summarises the main points
418-419 and 420-421 related to field survey pollution data: You do not have results to back up these conclusions.
***This has been removed.
***Thank you very much for your detailed comments and suggestions. We greatly appreciate the effort you have taken to review our manuscript and feel it is much stronger as a result
Reviewer 2 Report
Review of manuscript submitted to Sustainability by Hastings et al
The manuscript by Hastings and colleagues reports on a study that attempted to model the potential distribution of two species of seagrass in Ireland. Such studies can be useful, and the methods seem generally appropriate (although see comments below), but I was surprised that no substantial effort was made to validate model predictions, either through field visits or earth observation methods. As a result, the model outcomes are a modest contribution – we don’t know how much of the potential distribution is actually occupied by seagrass.
The authors invoke the potential for seagrass to sequester organic carbon as a rationale for the study, but their efforts are misdirected for several reasons. Most importantly, there are no estimates of carbon stock or sequestration rates, so any link to organic carbon is entirely speculative. This is made worse by the apparent poor understanding of the processes of carbon capture and sequestration. For example, seagrasses use multiple sources of inorganic carbon, especially bicarbonate, not just CO2 (line 45), sequestration of organic carbon by seagrasses is not a “CO2 stocking rate” (line 65), and so on. Unfortunately, the inaccuracies and poor framing detract from the manuscript.
On the method, bathymetry is used as a proxy for light availability, but this assumes that light penetration is constant throughout Ireland’s coastal waters. This seems unlikely. Since light is the most critical predictor for subtidal seagrass (unsurprisingly), it does mean that model predictions for Zostera marina are probably inaccurate.
The model also used a resolution of one kilometre. This is understandable based on the data available, but is very coarse for seagrass – meadows are often smaller than this – and so might generate underestimates (as their field surveys in Cork showed).
Finally, the authors speculate about the utility of the model for restoration. This is unjustified, because restoration implies that seagrass is returned to a place where it once existed, but exists no longer. The authors have not shown that this is the case for any location (if they could, it would certainly be valuable).
Overall, the model is a modest, but useful contribution. Ideally the authors would provide better validation, but the effort might be adequate without this if they can simplify the manuscript to remove the speculative and unjustified aspects relating to carbon storage and restoration.
Author Response
***Denotes author responses
The manuscript by Hastings and colleagues reports on a study that attempted to model the potential distribution of two species of seagrass in Ireland. Such studies can be useful, and the methods seem generally appropriate (although see comments below), but I was surprised that no substantial effort was made to validate model predictions, either through field visits or earth observation methods. As a result, the model outcomes are a modest contribution – we don’t know how much of the potential distribution is actually occupied by seagrass.
***Along with comments from reviewer 1, we have elaborated on why we could not ground truth this data in the discussion.
The authors invoke the potential for seagrass to sequester organic carbon as a rationale for the study, but their efforts are misdirected for several reasons. Most importantly, there are no estimates of carbon stock or sequestration rates, so any link to organic carbon is entirely speculative. This is made worse by the apparent poor understanding of the processes of carbon capture and sequestration. For example, seagrasses use multiple sources of inorganic carbon, especially bicarbonate, not just CO2 (line 45), sequestration of organic carbon by seagrasses is not a “CO2 stocking rate” (line 65), and so on. Unfortunately, the inaccuracies and poor framing detract from the manuscript.
***Thanks for catching. We have now gone back through the manuscript and ensured that we have removed any reference to CO2 to reflect this process.
On the method, bathymetry is used as a proxy for light availability, but this assumes that light penetration is constant throughout Ireland’s coastal waters. This seems unlikely. Since light is the most critical predictor for subtidal seagrass (unsurprisingly), it does mean that model predictions for Zostera marina are probably inaccurate.
***We have provided references that have undertaken similar work. We agree that there are assumptions in this, but alternative variables such as diffuse attenuation were not available at a suitable spatial scale. Subsequently, we now address in the discussion
The model also used a resolution of one kilometre. This is understandable based on the data available, but is very coarse for seagrass – meadows are often smaller than this – and so might generate underestimates (as their field surveys in Cork showed).
***Again, we have addressed this in the discussion, highlighting possible discrepancies between the scale the processes operate at.
Finally, the authors speculate about the utility of the model for restoration. This is unjustified, because restoration implies that seagrass is returned to a place where it once existed, but exists no longer. The authors have not shown that this is the case for any location (if they could, it would certainly be valuable).
***A good point. Following this comment, we did explore historical locations, but given the lack of validation data, we could not state with certainty. As such, we have just clarified this in the writing, and also introduced the idea of enhancement, which does not require previous presence.
Overall, the model is a modest, but useful contribution. Ideally the authors would provide better validation, but the effort might be adequate without this if they can simplify the manuscript to remove the speculative and unjustified aspects relating to carbon storage and restoration.
***Thank you for your comments and suggestions. We feel we have removed some of the speculative comments from the work, and as such it should now hopefully make a solid contribution
Round 2
Reviewer 1 Report
The authors have done a good job editing the manuscript based on previous reviewer comments. The most prominent issue that remains is the need to improve Figure 2 maps to better show which colors represent “suitable” vs. “non-suitable” habitat for seagrasses. Leaving the map as-is feels like a missed opportunity, as the figure could serve to summarize the findings of the modelling in the paper, but the color coding of the maps make it very difficult to interpret as of now. Additionally, minor changes needed relate to newly added text that needs clarification or re-wording. See specific comments below.
Line 32: ‘We identified presence of seagrass….” – not sure what this is referring to. Are you referring to the dataset of seagrass distribution you used or the suitable areas you identified? Reword?
Line 33: “although part of this reasoning…” – please clarify this sentence. Are you referring back to your finding in the previous sentence to justify why anthropogenic variables did come up as important in your models? This sentence needs reworking.
Line 220: Figure 2 – These maps are still difficult for distinguishing suitability level. As they are, it is clear what is higher suitability and lower, but that is all. Are the large areas of lighter shades of orange and blue “0” suitability or something higher? A color gradient using different colors for different levels of suitability (even just “suitable” vs. “unsuitable” as on line 230) would be easier to read. Example color schemes (including color-blind-friendly) can be seen here: https://colorbrewer2.org/#type=diverging&scheme=BrBG&n=4
Lines 253-260: This is a very long sentence. Can the second half of it be broken up and clarified? E.g. Line 256-257: Do you mean to say there are potentially large areas of unmapped seagrass at the national level?
Author Response
***Author responses
The authors have done a good job editing the manuscript based on previous reviewer comments. The most prominent issue that remains is the need to improve Figure 2 maps to better show which colors represent “suitable” vs. “non-suitable” habitat for seagrasses. Leaving the map as-is feels like a missed opportunity, as the figure could serve to summarize the findings of the modelling in the paper, but the color coding of the maps make it very difficult to interpret as of now. Additionally, minor changes needed relate to newly added text that needs clarification or re-wording. See specific comments below.
***Thanks for taking the time to review the manuscript so extensively again, and the positive comments
Line 32: ‘We identified presence of seagrass….” – not sure what this is referring to. Are you referring to the dataset of seagrass distribution you used or the suitable areas you identified? Reword?
***This should have been deleted with the previous edit. We have now removed this, and it should not read ambiguously anymore.
Line 33: “although part of this reasoning…” – please clarify this sentence. Are you referring back to your finding in the previous sentence to justify why anthropogenic variables did come up as important in your models? This sentence needs reworking.
***We have reworked this sentence to add clarity
Line 220: Figure 2 – These maps are still difficult for distinguishing suitability level. As they are, it is clear what is higher suitability and lower, but that is all. Are the large areas of lighter shades of orange and blue “0” suitability or something higher? A color gradient using different colors for different levels of suitability (even just “suitable” vs. “unsuitable” as on line 230) would be easier to read. Example color schemes (including color-blind-friendly) can be seen here: https://colorbrewer2.org/#type=diverging&scheme=BrBG&n=4
***We have now revised this figure to include low, moderate, high, and very high suitability as per your suggestions. Note the comments about the ambiguous nature of thresholding from the previous round of revisions persists, but this should hopefully rectify your concerns. We have used the colour scheme from the provided link.
Lines 253-260: This is a very long sentence. Can the second half of it be broken up and clarified? E.g. Line 256-257: Do you mean to say there are potentially large areas of unmapped seagrass at the national level?
***We have split this sentence and it should be clarified now.
***Thank you again for your constructive reviews
