Forest Landscape Heterogeneity Increases Shrub Diversity at the Expense of Tree Seedling Diversity in Temperate Mixedwood Forests
Round 1
Reviewer 1 Report
This study is essential for understanding of relations among forest layers and components and reveals the mechanism how the understory influences diversity of natural tree regeneration. I have no objection to this study, and I found no error. Maybe two comments: Authors deal with diversity in forests and gaps in forest. However, the gaps in forest encompassed only small proportion and this fact is also properly introduced in the study. So, the title of the study is Landscape heterogeneity …. but the title Forest heterogeneity…..expressed better the topic of the study.
Secondly authors formed stand types combining classes of forest overstory composition type, stand cover density and height classes. While the first two criteria are explained I do not know the criteria (limits) for height classes. The limits should be mentioned.
On the other hand, this study is only a part of mosaic which should reveal a dynamic of growth and survival of tree species during natural regeneration of forest stands. One must realize that there is difference between number of tree species and number of individuals within tree species growing under shrub layer. So, we need more studies dealing with relation of growth dynamics between shrub layer and tree regeneration for general recommendations for forest owners and managers.
Author Response
Dear Reviewer 1
We are heartened by your positive review. We have considered your suggested change to the title. We wanted to emphasise that structure is different across forested landscapes in other words we wanted to put the emphasis on the often-ignored landscape scale. Title “Landscape heterogeneity” changed to “Forest landscape heterogeneity.
We acknowledge the reviewer’s comments that landscape heterogeneity will manifest through not only diversity, but also abundance, growth and survival and be dynamic through time. We acknowledge the reviewer’s mention of the difference between number of species and number of individuals, and the interaction between these terms, which we deal with through rarefaction curves in Figures S1 and S2.
Change made to height classes: “and height (0-6 height class)” to “and height (seven classes based on the height of dominant et codominant trees of the stand (1: >22m; 2: (17m-22m); 3: (12m-17m); 4: (7m-12m); 5: (4m-7m); 6: (2m-4m); 7: <2m)”. L.130-131.
We would like to thank Reviewer 1 for taking the time to read and critique our manuscript,
Sincerely,
The Authors,
Reviewer 2 Report
Dear authors,
The study investigates the effect of landscape heterogeneity on shrub and tree species diversity in temperate mixed forests by considering alpha- and beta-diversity as well as different gap sizes. It is really a great opportunity to analyse these differently grained landscapes. Also, the idea of considering these two growth forms is quite valuable and can have important implications for forest management and particularly forest understorey management.
However, I am not very convinced about the main conclusion on direct competitive effects of shrubs on tree species diversity. The results you present give no direct evidence that heterogenous landscapes or the different gap types show a negative relationship between shrub abundance/shrub diversity and tree species diversity. Thus, the conclusion is just based on an assumption but not on real evidence. The lines 404-407 seem to rather represent the authors opinion than a conclusion based on the presented results. May it be possible to directly analyse the effect of shrubs on tree species diversity on the alpha level?
In addition, you only present species accumulation curves for the different landscapes separated by gap type (forest, small, medium, large). There is, however, no overall comparison between the landscapes regarding tree and shrub species diversity, in other words a comparison of gamma-diversity is missing. A landscape thereby should comprise forest understories as well as different gap types with heterogenous landscapes comprising a higher proportion of large gaps. It could be hypothesized that both for trees and for shrubs different habitat types harbour different species assemblages and therefore more heterogenous landscapes harbour more overall shrub and tree species. This aspect is missing in your study. As you have sampled 18 sites per landscape, you have a similar sampling effort in all 12 landscapes, and you can do a direct comparison of total gamma-diversity between the three landscape types.
Specific comments:
L99-123: In the study area section, you mention different natural or anthropogenic disturbances that have affected the forests in the past or are affecting the forests currently. I would appreciate a better connection between the assumed “natural landscape composition and configuration” and the three landscape heterogeneity categories that you have built. Is the homogenous landscape the „natural landscape“ and the heterogenous landscape is mainly created by anthropogenic disturbances?
L141: The heterogenous landscape should comprise index values of 0.62-0.77 as landscape 50 has a value of 0.61 and was assigned to the moderate category.
L150: Do you mean that 50 % should be mixed forests of these three tree species?
L159: From my point of view the term gap / forest environment is a bit misleading as you have not investigated a gradient, it is a gap type or gap size you have investigated.
L164-166: It would be helpful to present how many micro quadrats were sampled in total and per landscape type and gap type.
L185-187: I do not understand this paragraph. According to Section 2.2. you assigned the landscapes to heterogeneity levels based on the landscape heterogeneity index calculated based on four sub-indices. Thus, what kind of validation do you mean?
L197: It should be landscape instead of site. As I understand your study design you have sampled 12 landscapes and 18 sites within each landscape.
L198-200: Why haven’t you analysed beta-diversity between landscape heterogeneity levels? You have only analysed beta-diversity for shrubs and trees within different gap types within the different landscapes. I would assume a higher beta-diversity in heterogeneous landscapes if sampled sites across all gap types would have been considered. Nevertheless, the higher species richness of trees in forest understories in heterogeneous landscapes is quite interesting and indicates a positive effect of heterogenous environments even within rather undisturbed forests.
L203: What do you mean with standardized species abundance (0-1)? Is it presence/absence? You mention the abundance of species or species groups at times but there are no data on abundance presented at all.
L206-207: Species accumulation curves can also indicate landscape-level gamma diversity. In my opinion, the study would benefit from a focus on landscape-level diversity across gap types.
Table 2: In this table you present 4 different gap sizes, but you have only analysed gaps sizes > 40 m². What happened to the category 4-40 m²? Is this considered as forest understory?
L231-233: This is an important point. Gaps make up an average of 2.5 % in all landscapes; thus, can a higher shrub diversity in in medium sized gaps in heterogenous landscapes really effect tree species diversity in the landscape negatively even when this species groups benefits in forest understories? What about potential positive effects of a higher shrub diversity for example for other organismic groups (e.g. herbivores)?
L248: shade-tolerant
L257 (Figure 3): Viburnum lantanoides should be Vila
Table 4: I do not see the value of this Table. It would be better to present the Shannon and Simpson diversity indices with bar plots or box plots to have the actual values. The significant differences could be marked with different letters.
Figure 5: Figure 5 shows similar numbers than presented in Table 5. Therefore, I think that Figure5 is not necessary.
Figure 6: The different categories are hard to differentiate. Why do you not use different colours?
L310-311: This sentence only refers to Table 4. Again, it would be better to present the real diversity indices. The statistical values alone do not show very much.
Figure 7: The same as for Figure 6, the categories are difficult to differentiate without different colours.
L328: Where do these numbers come from? The alpha diversity values (neither species richness per 5 m² nor the diversity indices) are not shown.
L346-347: This is speculative. Is there a negative relationship between shrub abundance or diversity and tree species diversity?
L374-375: How is this sentence connected to your study? As you have an increase in tree species in forest understories in heterogenous landscapes, you cannot really talk about biodiversity loss. In addition, there are no overall numbers on species richness or species diversity for whole landscapes.
L379: You present no data on abundance.
L387-389: See above. You present no direct evidence that there is a competitive exclusion of tree species by shrubs.
L392-407: The conclusion is highly speculative and should be rewritten.
Figure S1 and S2: You should clarify what is meant with sample size. It is not self-explaining that it is the number of individuals (I assume).
Author Response
Dear Reviewer 2,
We’re encouraged by the reviewer’s mention of the great opportunity presented by our work and the value for forest management. We agree that a direct analysis of competition between shrubs and trees is not presented in this paper. Competition is an effect cited in the literature and our interpretation is based on this (Royo and Carson 2006). In response to the comment ”The results you present give no direct evidence that heterogenous landscapes or the different gap types show a negative relationship between shrub abundance/shrub diversity and tree species diversity.” And to ”May it be possible to directly analyse the effect of shrubs on tree species diversity on the alpha level?” we’ve added a simple regression to the results section to support this interpretation. We’ve also nuanced the text to recognize that it may be due to other factors such as differential colonisation.
We agree with the reviewer that a presentation of gamma-diversity would be a supportive element for the demonstration in our article. The reviewer suggested: ”As you have sampled 18 sites per landscape, you have a similar sampling effort in all 12 landscapes, and you can do a direct comparison of total gamma-diversity between the three landscape types”. We chose to use a stratified sampling strategy that specifically samples gaps to ensure we would have enough observations for the analyses (reduce zeros in the dataset) and control for light environment while looking into heterogeneity. Alternatively, a completely randomized sampling of sites would have given almost exclusively forest sites with much lower luminosity, and would have been appropriate to scale up to gamma level analyses. Site stratification (9 forest sites and 9 gap sites) does not reflect the natural distribution of gap and forest sites in the landscape (the main effect landscape heterogeneity). Comparing gamma diversity like the reviewer suggested does not allow taking into account the different proportions of the different forest environments and would cause a methodological problem.
Another reason that we do not discuss gamma diversity is that our sample size may not be large enough as our species accumulation curves have not attained saturation (in theory, we should keep on sampling until they attain saturation for comparing gamma diversity).
We would like to thank Reviewer 2 for taking the time to read and critique our manuscript. We are thankful for the reviewer’s critique which we feel has significantly improved the manuscript.
Sincerely,
The Authors,
Specific comments:
L99-123: In the study area section, you mention different natural or anthropogenic disturbances that have affected the forests in the past or are affecting the forests currently. I would appreciate a better connection between the assumed “natural landscape composition and configuration” and the three landscape heterogeneity categories that you have built. Is the homogenous landscape the “natural landscape“ and the heterogenous landscape is mainly created by anthropogenic disturbances?
Response: No not necessarily, there are anthropogenic disturbances across all landscapes, we’ve modified the text for clarity. L.153-155.
L141: The heterogenous landscape should comprise index values of 0.62-0.77 as landscape 50 has a value of 0.61 and was assigned to the moderate category.
Response: We’ve accepted the modification, thank you. L.144.
L150: Do you mean that 50 % should be mixed forests of these three tree species?
Response: No, we mean of that forest type, not necessarily those species. We’ve modified the text to add the words “forest type”. L.156.
L159: From my point of view the term gap / forest environment is a bit misleading as you have not investigated a gradient, it is a gap type or gap size you have investigated.
Response: This comment is appreciated as the authors have been debating over the proper term for some time now, and this comment will stimulate more discussion. We have looked at “habitat”, “gap / forest environment”, “light environment” and “gap size” as the effect name. It’s curious why gap-forest is seen as a gradient and not a dichotomy. After exchanging on your comment we decided to go keep the “gap / forest environment”.
L164-166: It would be helpful to present how many micro quadrats were sampled in total and per landscape type and gap type.
Response: We’ve added a table (Table S1) with the sum of microquadrats and gaps sampled per landscape type, gap type and interaction to supplementary materiel. L.203.
L185-187: I do not understand this paragraph. According to Section 2.2. you assigned the landscapes to heterogeneity levels based on the landscape heterogeneity index calculated based on four sub-indices. Thus, what kind of validation do you mean?
Response: We’ve modified the text “for validation” to “to examine how each sub-component differed among the three classes we generated using the global spatial landscape heterogeneity index”. L.195-196.
L197: It should be landscape instead of site. As I understand your study design you have sampled 12 landscapes and 18 sites within each landscape.
Response: We had three levels of stratification, we sampled in microquadrats, within sites, within landscapes. This has been changed to “and the site identifier (sampled site (1-223)) as a random factor”. L.206. The random factor is usually one spatial step above the scale of analysis. For ANOVAs we analyzed at microquadrat level, so the next step up is site identifier. If you analyze at the site level, then the random factor would be the landscape.
L198-200: Why haven’t you analysed beta-diversity between landscape heterogeneity levels? You have only analysed beta-diversity for shrubs and trees within different gap types within the different landscapes. I would assume a higher beta-diversity in heterogeneous landscapes if sampled sites across all gap types would have been considered. Nevertheless, the higher species richness of trees in forest understories in heterogeneous landscapes is quite interesting and indicates a positive effect of heterogenous environments even within rather undisturbed forests.
Response: We agree with the reviewer’s proposed hypothesis when combining levels, however we cannot combine levels for the sampling reasons mentioned above. We also agree with the reviewer’s interpretation of our results.
L203: What do you mean with standardized species abundance (0-1)? Is it presence/absence? You mention the abundance of species or species groups at times but there are no data on abundance presented at all.
Response: We’ve modified the text to “standardized frequency”. L.214.
L206-207: Species accumulation curves can also indicate landscape-level gamma diversity. In my opinion, the study would benefit from a focus on landscape-level diversity across gap types.
Response: We cannot combine levels for the multiple sampling reasons mentioned above. Moreover, as the SAC do not reach the asymptote (saturation), gamma diversity could not be extrapolated nor compared using the SAC.
Table 2: In this table you present 4 different gap sizes, but you have only analysed gaps sizes > 40 m². What happened to the category 4-40 m²? Is this considered as forest understory?
Response: Not by design. But for the very small ones, it is almost impossible to detect them from below (while sampling in the forest) and to avoid them completely; we acknowledge that some may have been included in the “forest understory” class.
L231-233: This is an important point. Gaps make up an average of 2.5 % in all landscapes; thus, can a higher shrub diversity in in medium sized gaps in heterogenous landscapes really effect tree species diversity in the landscape negatively even when this species groups benefits in forest understories? What about potential positive effects of a higher shrub diversity for example for other organismic groups (e.g. herbivores)?
Response: This comment brings up the relative contribution of gap dynamics to regenerate plant species, given that in our study the landscape is mostly forest, although interesting, it remains outside the scope of our paper. The reviewer also mentions interactions with other organisms, which is also very interesting, but outside the scope of this paper. Finally, the competitive effect of shrubs on tree is dealt with in the above comments.
L248: shade-tolerant
Response: Modification made L.264.
L257 (Figure 3): Viburnum lantanoides should be Vila
Response: We’ve kept the old latin name “Viburnum alnifolium” in the text.
Table 4: I do not see the value of this Table. It would be better to present the Shannon and Simpson diversity indices with bar plots or box plots to have the actual values. The significant differences could be marked with different letters.
Response: The table explains the differences between levels for main effects. We’ve received multiple comments that the article has too many figures. Although we acknowledge that the reviewer has a good point, we’ve kept the table instead of adding another figure.
Figure 5: Figure 5 shows similar numbers than presented in Table 5. Therefore, I think that Figure 5 is not necessary.
Response: We’ve accepted the modification, Figure 5 was moved to supplementary material. (Now Figure S1).
Figure 6: The different categories are hard to differentiate. Why do you not use different colours? (Now Figure 5).
Response: We’ve accepted the modification, we’ve added colour.
L310-311: This sentence only refers to Table 4. Again, it would be better to present the real diversity indices. The statistical values alone do not show very much.
Response: Changed to refer only to Table 4. L.337. Although we acknowledge that the reviewer has a good point, no change was made to add tables, we’ve received multiple comments that we have too many tables.
Figure 7: The same as for Figure 6, the categories are difficult to differentiate without different colours.
Response: We’ve accepted the modification, we’ve added colour. (Now Figure 6).
L328: Where do these numbers come from? The alpha diversity values (neither species richness per 5 m² nor the diversity indices) are not shown.
Response: Taken from the text L.252-257. Species richness HET=2.38 species/microquadrat, Mod=1.98 species/microquadrat, Hom=1.81 species/microquadrat. HET-MOD:2.38-1.98=0.40, HET-HOM: 2.38-1.81=0.57.
L346-347: This is speculative. Is there a negative relationship between shrub abundance or diversity and tree species diversity?
Response: We’ve accepted the suggestion and added “We found a negative relationship between shrub abundance and tree seedling diversity using simple regression (T value = -3.71, P(f) = <0.01)”. L.258-259.
L374-375: How is this sentence connected to your study? As you have an increase in tree species in forest understories in heterogenous landscapes, you cannot really talk about biodiversity loss. In addition, there are no overall numbers on species richness or species diversity for whole landscapes.
Response: We’ve accepted the suggestion and taken out the phrase “Hence, worldwide loss of specialist species and biodiversity due to increased disturbance and habitat destruction [62]”. L.400-401.
L379: You present no data on abundance.
Response: We’ve taken out this mention of abundance. L.405.
L387-389: See above. You present no direct evidence that there is a competitive exclusion of tree species by shrubs.
Response: We’ve accepted the suggestion and added a significant simple regression mentioned above.
L392-407: The conclusion is highly speculative and should be rewritten.
Response: We’ve accepted the suggestion and have changed the conclusion. The added significant linear regression mentioned above however comes to support the elements that have been put in the conclusion within, we believe, the scope and the interpretation limits of the study.
Figure S2 and S3: You should clarify what is meant with sample size. It is not self-explaining that it is the number of individuals (I assume).
Response: We’ve accepted the modification, we’ve modified the label to Sample size (number of individuals).
Reviewer 3 Report
MS Review: Landscape Heterogeneity Increases Shrub Diversity at the Expense of Tree Seedling Diversity in Temperate Mixedwood Forests
General comments
The manuscript “Landscape Heterogeneity Increases Shrub Diversity at the Expense of Tree Seedling Diversity in Temperate Mixedwood Forests” uses a combination of remote sensing and field sampling to assess the influence of landscape patterns in forest habitat structure (fragmentation and “heterogeneity”) on the composition, diversity, and structure of shrub and tree seedlings in mature and regenerating forest. Heterogeneity in structure at multiple scales from the stand to landscape is often promoted as a strategy for maintaining or restoring biodiversity, so the questions and associated research presented in this paper are relevant to forest science and have potential to influence forest management decisions.
In general, the manuscript is well written. I enjoyed reading the introduction and discussion sections, and while I have some questions and suggested in “specific comments” below, I generally found that the arguments presented were well supported. (This is frequently not the case, so I intend that as a complement!) I have two primary concerns. First, the study only includes saplings and shrubs with height > 20 cm and basal diameter < 1cm, assessed independently of surrounding, larger vegetation that undoubtedly has an impact on composition and abundance in the seedling layer. Second, some of the statistical approaches do not appear to have been as robust as they could be. (This could possibly be addressed with more detail, but I have questions about why some of the approaches were used.) I also recommend some discussion of how heterogeneity and fragmentation relate. Heterogeneity is often promoted as a good thing in forest ecosystems, depending on context and scale, while fragmentation is widely recognized to have negative consequences. I may have misunderstood, but as I interpret the methods, greater fragmentation at the landscape scale would lead to classification of greater heterogeneity in this study.
Specific comments
Lines 26-28: I recommend considering the removal of the statements about rate of species accumulation, depending on how later comments (referencing lines 205-218) related to the analysis and interpretation are addressed.
Lines 54-56: If possible, you might consider breaking this sentence (describing mechanism 3) up. I realize that is challenging given the list structure used here, but as written, it is a little difficult to follow.
Lines 59-62: References 11, 12, 14, and 15 (the bulk of the support for this paragraph) all describe studies in agricultural landscapes. Heterogeneity in a landscape dominated by natural ecosystems is not likely to be the same as heterogeneity in an agricultural landscape with natural remnants. The relevance of the cited studies to the forested landscape studied here needs to be made clearer. It is sensible to refer to those studies and explain the importance of knowing whether similar trends occur in a landscape dominated by natural ecosystems (with heterogeneity/fragmentation resulting from different mechanisms), but some distinction needs to be drawn.
Line 74: This line was confusing. Consider changing “dominate” to “limit”, “compete with”, or something similar.
Lines 99-123: The study area description was well done. I appreciate the detail included.
Line 120: The word “nowadays” is colloquial and a little informal for this outlet. I recommend something like “increasingly” or “in recent years” instead.
Lines 125-141: Was forest cover equivalent among the different heterogeneity classes? This seems like a key assumption that should have been controlled for, but it is not mentioned until the results section (lines 232-233). If forest cover (and by default canopy disturbance/gap cover) is not consistent across heterogeneity classes then more heterogenous areas may simply be more disturbed.
Line 126: Are stand and site synonymous? Based on the terminology used later in the paper, I think it may be most appropriate to call the polygons “stands”. This needs clarification early, and then everything that follows should be consistent.
Figure 1: What do the grey and white shading represent in panel a? This is indicated for panel b, but I don’t see any explanation for a. Clearly there is some pattern across the landscape, and it would be helpful to know what it is…
Table 1: Should the lower-case letter “a” in superscript for mean Shannon diversity in “homogenous” landscapes actually be “b”? It looks more similar to the mean Shannon diversity in “moderate” landscapes, but the current designator indicates it is statistically indistinguishable from “heterogeneous” landscapes.
Table 1: The information presented in Tables 1, 2, & 5 would be easier to interpret (and more impactful) if presented as bar graphs with error bars instead.
Table 1: If the table is retained, I recommend adding lower case letters that indicate differences (or an absence of difference) for all variables rather than only those where statistical tests indicated a difference. This is a less confusing approach and confirms that statistics were actually performed.
Line 240: Recommend changing “inversely” to “In contrast”.
Line 156: It would be more appropriate (and less confusing) if “plot” were used here instead of “site”. Later, sites are defined as the 12 sampled landscapes.
Lines 167-169: It is extremely unfortunate that only seedlings are included, especially considering the emphasis on scale. The abundance of saplings and larger shrubs (diameter at 10 cm > 1 cm) would have an enormous impact on seedling densities, and I am not convinced that they can be interpreted independently.
Line 189: R is referenced in line 218, but it should really be mentioned each time specific packages are named or at least early on in the methods rather than at the end.
Lines 203-205: Why wasn’t mixed effects ANOVA used to test for differences in the spatial median (distance between plots and the centroid in multivariate space)? This would be a more robust approach than relying completely on pairwise comparisons. Plus, ANOVA allows a direct test of the potential interaction landscape heterogeneity and local gap/forest environment. If ANOVA indicates differences among groups, pairwise comparisons can then be used to determine which groups differ from one another.
Lines 206-218: Scheiner (2004) is cited here, but I don’t see any indication of the “type” of species-accumulation curve (Type II?) being presented. This is worth mentioning briefly because it influences interpretation. This also relates to some of the comments I made about the description of heterogeneity in the introduction – all of the references cited describe studies from agricultural landscapes (more likely Type IV curves where a forest remnant might be considered an island) whereas the heterogeneity and interaction between and among habitat types in a mostly forested landscape would be expected to differ.
Lines 207-208: I may have misunderstood what is being described in this statement, but I suggest being cautious about interpreting the ratio of species to number of individuals sampled (slope of the species-accumulation curve). It can be misleading. See Gotelli and Colwell (2001). Scheiner (2004) addresses this at least partly when correcting his own earlier paper (Scheiner 2003) and includes the intercept (in addition to the slope) for the curve as a measure of beta diversity.
Lines 211-212: Rarefaction analysis can be done based on samples or individuals. In fact, I suspect a sample-based rarefaction may be more appropriate in this case, even though accounting for the problem of varying densities (identified by the authors) requires extra effort. Gotelli and Colwell (2011, p.46-47) provide recommendations and examples for doing this. They (and others) also point out that a key assumption of rarefaction is that comparable sampling methods are used for the groups being compared.
Line 217: How was the confidence interval calculated? See Gotelli and Colwell (2011, p. 47).
Table 2: I recommend changing the caption so that it says something like “Proportion of landscape in gaps of different size classes in the 12 sampled landscapes”. The original phrasing led me to expect the columns for each landscape to sum to 100%.
Figures 3 and 4: Recommendation: Indicate significant differences with lower-case letters and add error bars.
Table 3: Include degrees of freedom.
Table 4: Caption is confusing – what is meant by “specific hypotheses”? What statistical test was used? Why are all values for the Simpson Index “N/A”? Consider removing the columns for the Simpson index.
Line 277: Is this statement comparing beta diversity between shrubs and seedlings? The language is a little unclear, plus I do not see results that indicate a statistical test of differences between the two groups.
Table 5: I am not convinced that Tukey’s HSD was the best approach for testing for differences in beta diversity (as quantified using distances to centroid), but it if is retained in the manuscript, the authors might consider indicating significance in Figure 5 and removing Table 5.
Figures 6 & 7: The lines for the curves are somewhat difficult to see. If possible, make them a little thicker and potentially test different patterns for the CI lines to improve visibility.
Author Response
Dear Reviewer 3,
We’re encouraged by the reviewer’s mention of enjoying the introduction and discussion. A small specification regarding the first concern of the reviewer, we recorded shrub seedlings and saplings, this has been added to the section 2.3 Field sampling “We recorded shrub total abundance (seedling: height ≥ 20 cm, diameter at 10 cm height < 1 cm, and sapling: diameter at 10 cm height ≥ 1.1 cm and ≤ 9 cm) and tree seedling (height ≥ 20 cm, diameter at 10 cm height < 1 cm) abundance by species in each microquadrat”. L.174-175. We do consider light attenuation by larger shrubs and trees in the overstory as represented by the gap / forest main effect. We give greater detail concerning the reviewer’s comments on statistics below and can elaborate on any of the reviewer’s concerns.
We agree with the reviewer’s interpretation of heterogeneity and fragmentation. We consider fragmentation as encompassing studies with an inhospitable matrix, whilst landscape heterogeneity, the studies have an (at least somewhat) hospitable matrix. This is our cleavage that we chose to respect concerning the literature and our interpretation.
We would like to thank Reviewer 3 for taking the time to read and critique our manuscript. We are thankful for the reviewer’s critique which we feel has significantly improved the manuscript.
Sincerely,
The Authors,
Specific comments
Lines 26-28: I recommend considering the removal of the statements about rate of species accumulation, depending on how later comments (referencing lines 205-218) related to the analysis and interpretation are addressed.
Response: Consideration warranted. We interpret species-area curves as showing rate of species accumulation.
Lines 54-56: If possible, you might consider breaking this sentence (describing mechanism 3) up. I realize that is challenging given the list structure used here, but as written, it is a little difficult to follow.
Response: We attempted breaking it up but found the text more confusing. So no change was made.
Lines 59-62: References 11, 12, 14, and 15 (the bulk of the support for this paragraph) all describe studies in agricultural landscapes. Heterogeneity in a landscape dominated by natural ecosystems is not likely to be the same as heterogeneity in an agricultural landscape with natural remnants. The relevance of the cited studies to the forested landscape studied here needs to be made clearer. It is sensible to refer to those studies and explain the importance of knowing whether similar trends occur in a landscape dominated by natural ecosystems (with heterogeneity/fragmentation resulting from different mechanisms), but some distinction needs to be drawn.
Response: Consideration warranted. See above comments on landscape heterogeneity / fragmentation debate.
Line 74: This line was confusing. Consider changing “dominate” to “limit”, “compete with”, or something similar.
Response: We’ve accepted the modification. L.74.
Lines 99-123: The study area description was well done. I appreciate the detail included.
Response: Thank you.
Line 120: The word “nowadays” is colloquial and a little informal for this outlet. I recommend something like “increasingly” or “in recent years” instead.
Response: We’ve accepted the modification. L.120.
Lines 125-141: Was forest cover equivalent among the different heterogeneity classes? This seems like a key assumption that should have been controlled for, but it is not mentioned until the results section (lines 232-233). If forest cover (and by default canopy disturbance/gap cover) is not consistent across heterogeneity classes then more heterogenous areas may simply be more disturbed.
Response: Landscape selection in the methodology describes the care we took in selecting landscapes with comparable abiotic and biotic features (Lines 151-159). They were all forest-dominated landscapes as demonstrated in Table 2. Also, in Table 2 we see that total gap area is not significantly different between landscape heterogeneity levels.
Line 126: Are stand and site synonymous? Based on the terminology used later in the paper, I think it may be most appropriate to call the polygons “stands”. This needs clarification early, and then everything that follows should be consistent.
Response: We’ve accepted the modification and removed sites from line 127. We agree with the reviewers’ interpretation and use sites as the experimental unit and stands as polygons.
Figure 1: What do the grey and white shading represent in panel a? This is indicated for panel b, but I don’t see any explanation for a. Clearly there is some pattern across the landscape, and it would be helpful to know what it is…
Response: We’ve accepted the modification and we added “grey areas are waterbodies and waterways” to the label of the figure. L.145-146.
Table 1: Should the lower-case letter “a” in superscript for mean Shannon diversity in “homogenous” landscapes actually be “b”? It looks more similar to the mean Shannon diversity in “moderate” landscapes, but the current designator indicates it is statistically indistinguishable from “heterogeneous” landscapes.
Response: We’ve accepted the modification and changed the superscript to b, thank you. L.149.
Table 1: The information presented in Tables 1, 2, & 5 would be easier to interpret (and more impactful) if presented as bar graphs with error bars instead.
Response: Table 5 is presented as boxplots in Figure 5 (now S1). Although we acknowledge that the reviewer has a good point, no change was made to add figures since we’ve received multiple comments that the article has too many figures.
Table 1: If the table is retained, I recommend adding lower case letters that indicate differences (or an absence of difference) for all variables rather than only those where statistical tests indicated a difference. This is a less confusing approach and confirms that statistics were actually performed.
Response: We’ve accepted the modification and added superscript a`s to Figure 1. L.149.
Line 240: Recommend changing “inversely” to “In contrast”.
Response: We’ve accepted the modification. L.254.
Line 156: It would be more appropriate (and less confusing) if “plot” were used here instead of “site”. Later, sites are defined as the 12 sampled landscapes.
Response: This was a mistake, in that later situation it should be 223 sampled sites. L. 206.
Lines 167-169: It is extremely unfortunate that only seedlings are included, especially considering the emphasis on scale. The abundance of saplings and larger shrubs (diameter at 10 cm > 1 cm) would have an enormous impact on seedling densities, and I am not convinced that they can be interpreted independently.
Response: We respond to this criticism above. Our study focused on forest tree regeneration diversity, in response to shrub diversity. A wide range of sizes was already covered by the size limits we used. For the shrubs, it included the vast majority of the individuals. Larger individuals that overtop tree seedlings or shrubs of the study size were considered as a part of the “habitat” effect (gap " forest environment 4 classes).
Line 189: R is referenced in line 218, but it should really be mentioned each time specific packages are named or at least early on in the methods rather than at the end.
Response: No change made.
Lines 203-205: Why wasn’t mixed effects ANOVA used to test for differences in the spatial median (distance between plots and the centroid in multivariate space)? This would be a more robust approach than relying completely on pairwise comparisons. Plus, ANOVA allows a direct test of the potential interaction landscape heterogeneity and local gap/forest environment. If ANOVA indicates differences among groups, pairwise comparisons can then be used to determine which groups differ from one another.
Response: Because we used a stratified sampling, we could not test the full factorial (Landscape heterogeneity (LH) + Gap/forest environment (GFE) + LH*GFE) for beta-diversity. Beta and gamma diversity cannot be scaled-up for characterising landscapes by combining plots if plots are not sampled in proportion of their representativeness in the landscape. This is why all beta analysis are done by gap/forest environment separately.
We changed:
“Tukey’s “Honest Significant Difference” method was used to identify differences in β- diversity between pair-wise comparisons of the three landscape heterogeneity levels.”
To: “When homogeneity of multivariate dispersions was infirmed, Tukey’s “Honest Significant Difference” method was used to identify differences in β- diversity between pair-wise comparisons of the three landscape heterogeneity levels.” L.215-218.
Lines 206-218: Scheiner (2004) is cited here, but I don’t see any indication of the “type” of species-accumulation curve (Type II?) being presented. This is worth mentioning briefly because it influences interpretation. This also relates to some of the comments I made about the description of heterogeneity in the introduction – all of the references cited describe studies from agricultural landscapes (more likely Type IV curves where a forest remnant might be considered an island) whereas the heterogeneity and interaction between and among habitat types in a mostly forested landscape would be expected to differ.
Response: Given the debate on how to classify species-area curves between Gray et al. (2004), and Scheiner (2004) we prefer not to mention specifically which type of curve we present.
Lines 207-208: I may have misunderstood what is being described in this statement, but I suggest being cautious about interpreting the ratio of species to number of individuals sampled (slope of the species-accumulation curve). It can be misleading. See Gotelli and Colwell (2001). Scheiner (2004) addresses this at least partly when correcting his own earlier paper (Scheiner 2003) and includes the intercept (in addition to the slope) for the curve as a measure of beta diversity.
Response: We appreciate the reviewer’s concern for this interpretation of beta diversity and this is in part why we take the time to also measure beta diversity using betadisper multivariance. Scheiner (2003) proposes that beta-diversity equals “the extent to which species composition changes from one place to another”, “If an exponential or power function is fit to a Type 1 curve, the intercept (c) measures alpha diversity, and the slope (z) measures beta diversity”.
As the reviewer mentions, Scheiner (2004) offers some specification on the original idea “I admit that the precise way that I presented this concept in Scheiner (2003) was in error. I stated that, ‘If an exponential or power function is fit to a Type I curve, the intercept (c) measures α diversity, and the slope (z) measures β diversity’ (Scheiner, 2003; p. 444). My error was that the rate of rise of a power or exponential curve is a function of both c and z (White & Gould, 1965; Connor & McCoy, 1979; Rosenzweig, 1995). Although my precise statement was wrong, the concepts were correct: α diversity can be measured as the height of the curve at a given area and β diversity can be measured as the rate of rise of the curve.”
Our curves use a power function and we apply this interpretation of beta diversity to species-area curves by comparing the lines which include consideration of slope and confidence intervals. We do not see direct contradiction with this interpretation in Gotelli and Colwell (2001, 2011). In Gotelli (2001) Figure 4 is referring to a problem that can be corrected by comparing the slopes of the lines at the same sample size (number of individuals).
Lines 211-212: Rarefaction analysis can be done based on samples or individuals. In fact, I suspect a sample-based rarefaction may be more appropriate in this case, even though accounting for the problem of varying densities (identified by the authors) requires extra effort. Gotelli and Colwell (2011, p.46-47) provide recommendations and examples for doing this. They (and others) also point out that a key assumption of rarefaction is that comparable sampling methods are used for the groups being compared.
Response: We’ve added Gotelli and Colwell (2011) as a citation [50]. Using Gotelli and Colwell (2011) terminology we used both sample-based rarefaction (Figures 5 and 6) and individual-based rarefaction (Figures S2 and S3) and we compared the outcomes of these species accumulation curves by gap-forest environment – landscape heterogeneity interaction to provide proof of direct diversity – heterogeneity effect.
In our article, site stratification (9 forest sites and 9 gap sites) that does not reflect the natural distribution of gap and forest sites in the landscape needs to be considered. This implication means we cannot analyse landscape heterogeneity species accumulation curves alone. To remedy the situation, the species accumulation curves (Figure 5 and S2) and dispersion analysis use the interaction between gap-forest environment – landscape heterogeneity thereby comparing heterogeneity levels inside of a specific light environment (with a similar amount of microquadrats and equal representation across the landscape, refer to Table S1 for number of microquadrats and sites per main effect and interaction).
Line 217: How was the confidence interval calculated? See Gotelli and Colwell (2011, p. 47).
Response: No comment.
Table 2: I recommend changing the caption so that it says something like “Proportion of landscape in gaps of different size classes in the 12 sampled landscapes”. The original phrasing led me to expect the columns for each landscape to sum to 100%.
Response: We’ve accepted the modification. L.233.
Figures 3 and 4: Recommendation: Indicate significant differences with lower-case letters and add error bars.
Response: No statistics were performed on frequency values. We present the frequency to give readers an idea of the regeneration in the forests we have studied. Error bars have been added.
Table 3: Include degrees of freedom.
Response: We’ve accepted the modification. L.296.
Table 4: Caption is confusing – what is meant by “specific hypotheses”? What statistical test was used? Why are all values for the Simpson Index “N/A”? Consider removing the columns for the Simpson index.
Response: We added “Tukey tests were used to evaluate the probability that categories relevant to our hypotheses were different”. L. 207-208. We removed Simpson values. L.295.
Line 277: Is this statement comparing beta diversity between shrubs and seedlings? The language is a little unclear, plus I do not see results that indicate a statistical test of differences between the two groups.
Response: No, this statement refers to shrubs only. To clarify we changed the first three sentences from
”β- diversity, as measured by multivariate dispersion, was generally higher in heterogeneous landscapes for shrubs for all gap / forest environments (Table 5). β- diversity was significantly different among landscape heterogeneity levels in the forest understory environment for the shrub species group (F value = 6.11, P(f) < 0.01) but not for any other gap / forest environment. In the forest understory, shrub β- diversity in homogenous landscapes was lower than in heterogeneous landscapes (Table 5, Figure S1).” to
“β- diversity, as measured by multivariate dispersion, was significantly different among landscape heterogeneity levels in the forest understory environment for the shrub species group (F value = 6.11, P(f) < 0.002) but not for any other gap / forest environment and was higher in heterogeneous landscapes than in homogeneous landscapes (Table 5, Figure S1).”L.297-305
Table 5: I am not convinced that Tukey’s HSD was the best approach for testing for differences in beta diversity (as quantified using distances to centroid), but it if is retained in the manuscript, the authors might consider indicating significance in Figure 5 and removing Table 5.
Response: Figure 5 has been moved to supplementary material (Figure S1). We added at the end of the caption of Table 5 “Values with different letters differ significantly according to the Tukey’s “Honest Significant Difference” method.”
Figures 6 & 7: The lines for the curves are somewhat difficult to see. If possible, make them a little thicker and potentially test different patterns for the CI lines to improve visibility.
Response: We’ve accepted the modification, we’ve added thicker lines and thicker confidence intervals. The figures are easier to read, thank you for this suggestion.
Round 2
Reviewer 2 Report
The authors responded well to all my comments and very well explained their decisions. Maybe the studied landscapes and the already compiled as well as future data may have the potential to answer other questions in terms of the effect of landscape heterogeneity on gamma-diversity of tree species as well as on other organismic groups in the future.
Thank you also for now providing a more direct evidence for an effect of shrub abundance on tree seedling diversity. I support publication of this paper.
