Population Density and Diversity of Millipedes in Four Habitat Classes: Comparison Concerning Vegetation Type and Soil Characteristics

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Statistical analysis is good, however, the Shannon's Index must be computed using logarithms in base 2. Authors used the standard logarithm (base 10); this procedure mae hard to compare results with other researchs. Please, correct the calculations. 

Changing the base of the logarithms will not affect the statistical results, but will improve the clarity of the manuscript.

The conclusions are somewhat tentative. In my opinion, the results allow for more robust conclusions, for example, mentioning which families are indicative of each type of use. And, above all, placing greater emphasis on how much density and richness decrease in agricultural ecosystems.

Author Response

• Statistical analysis is good, however, the Shannon's Index must be computed using logarithms in base 2. Authors used the standard logarithm (base 10); this procedure mae hard to compare results with other researchs. Please, correct the calculations. Changing the base of the logarithms will not affect the statistical results, but will improve the clarity of the manuscript.

Done. All the Shannon analyses were redone using log₂. The comparative results remain the same, although the values have changed. The necessary modifications were made in the main text, in the Appendices, and in the Supplementary Material.

 

• The conclusions are somewhat tentative. In my opinion, the results allow for more robust conclusions, for example, mentioning which families are indicative of each type of use. And, above all, placing greater emphasis on how much density and richness decrease in agricultural ecosystems.

Done. We have added the information you rightly suggested to the Conclusions.

Thank you very much

Reviewer 2 Report

Comments and Suggestions for Authors

General remarks

The authors conducted research on millipedes in the Valle Nuevo National Park, studying four different types of habitats. The effort and data obtained seem fairly typical for millipede research. Despite the undoubted value of the work for understanding the fauna of the Dominican Republic, it raises a number of questions. Firstly, the authors largely copied their own research from 10 years ago, obtaining very similar results. It is therefore difficult to speak of the scientific novelty of this research. They used a different methodology and collected data at a different time of year, so the results obtained cannot be compared with the past (because the season is different now than it was 10 years ago) or with other seasons (because the methodology is different now than it was 10 years ago). Secondly, it is unclear why these four types of environments were chosen for the study. Was the aim of the study to compare any environments, or was there a deeper idea or hypothesis behind the researcher's choice? Thirdly, the research hypotheses seem trivial. Differences in the occurrence of biota in such diverse environments seem obvious. Some of the methods used in many places are somewhat outdated, although still popular in millipede research. The use of multivariate methods would help to better analyze the collected material. I suggest rethinking the purpose of the research and hypotheses in order to find a deeper meaning in the research conducted, and then conducting the appropriate analysis again.

 

Detailed remarks

Throughout the text. Authors use interchangeably the terms groups, habitat classes, habitat types, ecosystem types, and sites. Please, unify nomenclature.

Line 12-14. The first two sentences of the abstract do not reveal any substantial information on the study results. It can be omitted.

Line 26-27. This statement does not follow directly from the research conducted.

Line 72-73. This hypothesis seems obvious; various habitats are inhabited by different assemblages of species. Especially when tree-covered ecosystems are compared to open ones.

Line 73-75. Statistical correlation is a tool in scientific research; the aim of the study is to find biological meaning using various methods, not the mathematical relation.

Line 90-91. It is not clear in what sense the term connectivity is used. Do authors mean connectivity in landscape, similarity of assemblages, or something else?

Methods section. It is not clear whether the samples of Millipedes were taken three times(November 21–24, 2020; December 7–8, 2020; and January 18-22, 2021) from the same 1x1 m quadrats. If so, I wonder how previous examinations influence the diversity and abundance in subsequent ones.

Line 185. Are you sure it's p greater than or equal to 0.05?

Line 192-197 and 205-213. It is not clear whether the diversity indices were calculated for the sample, the transect, or the site type. The procedure of Hutcheson and the Satterthwaite approximation should be described in more detail, or the reference should be given, since it is not a widely known procedure.

Line 222-230. There is an inconsistency in the way the results are presented, for TOV Pinoc and TUV Boslat - min, average, SE; for TOV and TUV Sabapa - max, average, SE; for BGT and RH - range, average, SE and CV. It is much better to use one scheme the data is presented.

Line 232 – What statistical test are the authors referring to?

Line 240-247. Are the differences in Millipede’s density and species richness statistically significant? This is much more important than deliberations on data dispersion. Try ANOVA or Kruskal-Wallis test.

Figures 3 and 4. The legend is unnecessary, but the axes description is mandatory. Please, add the result of the statistical analyses.

Line 275-276. It is an interpretation, not a result.

Figure 5. The figure is difficult to follow, especially because of the lack of description of the axes. A simple table would be more readable, in my opinion.

Figure 6. As above,  the figure is difficult to follow. A table with species in rows and habitat types in columns, with the number of each species in the table body, would give more information about shared species and their distribution among sites

Line 300-313. There are some mental shortcuts in the paragraph. What is the “quantitative distribution”? The lines 300-304 are not statistically precise. ANOVA reveals the differences among groups, and one statistically different value from the others is enough to find the result significant. To find which values are different, a post hoc test is required.  There is no information on the post hoc test used. The abundance of species in samples often deviates from normality, and the condition of homogeneity of variance is violated. The is no test presented that the conditions for ANOVA have been met.

Line 342-346; 350-354. Those explanations refer to the mathematical construction of the indices but have no biological meaning. With the table suggested in the remark for Fig. 6, for those who know the construction of the indices, it would be obvious without the additional explanations.

Table 3. The species richness should be presented in the table.

Line 375 -376; 379-378. As in remark for Line 342-346; 350-354, those explanations refer to the mathematical construction of the indices but have no biological meaning.

Line 380-383. It is an interpretation, not a result. Furthermore, it is not clear what the ecological connectivity or connection is.

Table 4. Why do authors not use ANOVA or the Kruskal-Wallis test for comparison of the Shannon diversity index, depending on the data characteristics, and then a post hoc test?

Table 5. I suggest using NMDS to show similarity relations among samples of the habitat type studied. I would reveal the similarity between habitat types and the variability of the species composition in each habitat type as well.

Line 400-404; 411-414. It is an interpretation, not a result.

Line 410, 411, 415,423,431. Those lines refer to the literature sources. In the result section, only the own results can be presented, not the literature references.

Line 432 -444. The approach of calculating the correlation of each environmental variable with species variables separately, in my opinion, is outdated. Modern analyses can analyze all environmental data and the abundance of each species and ordinate species, sites, and environmental variables on one graph, with statistical tests.  It allows for the finding of correlations between environmental variables, avoiding collinearity, and finding more universal patterns in the data. CCA or RDA with environmental variables selection seams good point to start.

Author Response

First of all, we thank the reviewer for their valuable remarks, which have helped us improve the quality of our manuscript.

 

• The authors conducted research on millipedes in the Valle Nuevo National Park, studying four different types of habitats. The effort and data obtained seem fairly typical for millipede research. Despite the undoubted value of the work for understanding the fauna of the Dominican Republic, it raises a number of questions. Firstly, the authors largely copied their own research from 10 years ago, obtaining very similar results. It is therefore difficult to speak of the scientific novelty of this research. They used a different methodology and collected data at a different time of year, so the results obtained cannot be compared with the past (because the season is different now than it was 10 years ago) or with other seasons (because the methodology is different now than it was 10 years ago).

 

You state that the authors have "copied" your research from 10 years ago. We must respectfully disagree, as this assertion is not accurate.

The study you refer to was an undergraduate thesis by a student of the first author. Based on that thesis, an article was later published, which we have cited in the Discussion section as a relevant precedent that should be acknowledged. Our current research, however, is original work that includes a substantial field component. Field surveys were conducted between July and August 2020, and the sampling of diplopods and soil took place between November and December 2020, and January 2021.

The article we submitted to the journal Ecologies presents partial results from a broader study that includes not only diplopods but also various groups of soil macroinvertebrates and their relationship with the physicochemical properties of the soil. This broader study will result in additional publications.

 

The similarities with the study from 10 years ago lie in the fact that both were conducted in the same region of Valle Nuevo National Park (PNVN), involved three of the four habitat classes included in our study, and used some commonly applied biodiversity indices for site comparison (Shannon, Margalef, and Jaccard).

Below, we highlight some of the many differences:

• Our research was conducted during the dry season in the highlands, whereas the previous study took place during the rainy season. Invertebrate populations in the high mountains of the Dominican Republic are known to be sensitive to seasonal changes. This has been documented not only for diplopods, but also for spiders, amphibians, and reptiles. Furthermore, the incidence of forest fires during the dry season plays an additional role. This factor is especially relevant for organisms with limited mobility, such as diplopods, which do not fly and move very slowly. Therefore, seasonality is not a minor or incidental variable—it is a critical factor that must be taken into account.
• Our comparison is indirect, and as you correctly noted, the sampling methodology and unit of effort differ. We will revisit that section of the text to ensure it cannot be misinterpreted and will make corrections if necessary.
• We employed a different methodology from the one used in the earlier study. This was not a minor adjustment, but a deliberate decision to streamline sampling under the extremely challenging conditions of the study area, and with limited resources. The goal was not to compare sampling techniques (which could be the focus of a separate study), but rather to apply a method that has proven effective for other zoological groups in the region, allowing for greater speed and efficiency. We also aimed to begin understanding seasonal dynamics in these populations.
• The scope of the current research is incomparably broader, as it includes a wide range of macroinvertebrates (arachnids, insects, annelids, mollusks, etc.), not just diplopods. This material will be used in future publications. These other soil fauna groups were not included in the previous study.
• Another indication of the broader scope is the inclusion of the agricultural ecosystem (Ecoag) among the habitat classes compared.
• Our article also includes an analysis of a wide array of soil physicochemical variables and determines their correlation with diplopod richness and diversity.
• The current study applies a much more comprehensive and rigorous statistical approach. We determine statistical significance in each case: normality tests are conducted before selecting the appropriate correlation coefficients when analyzing relationships between population density/richness and soil physicochemical variables. We also perform analysis of variance (ANOVA) to compare species distribution across the four habitat classes, as well as hypothesis tests using the Shannon diversity index to assess whether the values differ significantly between each pair of habitats. None of this statistical treatment was included in the earlier study.
• A careful comparison of our article with the one from 10 years ago clearly shows that the results differ even in comparable aspects: species richness, Shannon diversity index values between habitat types, similarity/dissimilarity among habitats, and population densities across the habitats are all different. Our article presents results and offers considerations that not only complement the earlier research with new elements but also expand upon it significantly.
• If we were to consider shared locations and habitat types as grounds for dismissing the novelty of a research project in terms of its approach, methodology, and contributions, then a vast number of studies published in scientific journals worldwide would also need to be disqualified. This line of reasoning would have to extend to the study of repeated research topics and objects, such as: biodiversity surveys involving tagging and organism counts; recurring use of standard statistics for alpha and beta diversity (sometimes presented in complex forms but fundamentally similar); studies on reproductive behavior through capture-recapture methods; construction of cladograms and phylogenetic trees; thousands of articles using just a handful of DNA techniques; and widely explored global topics like climate change, ocean warming, and ecosystem pollution, which often rely on common methodological tools.
• Numerous ecosystems—lakes, lagoons, deserts, seas, and many others—have been studied repeatedly, and that does not invalidate the merit of the new studies conducted in them.
• We believe the distinction between “what is novel” and “what is new” is well understood in scientific and cultural literature. Based on this understanding, we must acknowledge that only a very small fraction of the thousands of articles published daily present something entirely new in the strictest sense.
• Our research is not "new" in that absolute sense—and we do not claim it to be—but it does offer meaningful insights and an uncommon methodological approach in studies on millipedes. Specifically, our study includes: comparative variance analysis of species across sites for characterization; determination of species’ indicator value for each ecosystem or habitat class; hypothesis testing using the Shannon diversity index between habitat classes; and correlation analyses (preceded by normality testing) to explore relationships between population density, species richness, and 19 different soil variables.

However, you are right to criticize the inappropriate comparisons made in the text between research results obtained using different methodologies and objectives. We have made the necessary corrections.

 

• Secondly, it is unclear why these four types of environments were chosen for the study. Was the aim of the study to compare any environments, or was there a deeper idea or hypothesis behind the researcher's choice?

 

These ecosystems or habitat classes are located along a transect of the road within this region of the park. Riparian vegetation is also present in the immediate vicinity of the rivers; however, this environment is unproductive or unsuitable for a study focused on millipedes. The interest in Valle Nuevo National Park stemmed from the implementation of a larger project underway at the time, which provided access to the necessary resources for fieldwork and laboratory analysis. Additionally, prior knowledge of millipedes and other zoological groups in the area gave us confidence in the presence of a fauna suitable for a study of this nature.

 

• Thirdly, the research hypotheses seem trivial. Differences in the occurrence of biota in such diverse environments seem obvious.

 

We respectfully disagree with your opinion; the answer can be found in the Detailed Remarks

 

• Some of the methods used in many places are somewhat outdated, although still popular in millipede research.

 

In the article, we mention the scarcity of ecological studies on this group. This is precisely our aim: to strengthen the ecological approach in research on this group, which has been heavily biased toward systematics and taxonomy.

 

• The use of multivariate methods would help to better analyze the collected material.

 

We performed NMDS, Principal Component Analysis (PCA), and Multidimensional Scaling Analysis (see the final version of the manuscript)

 

• I suggest rethinking the purpose of the research and hypotheses in order to find a deeper meaning in the research conducted, and then conducting the appropriate analysis again.

 

Please see the answer above

 

Detailed remarks

• Throughout the text. Authors use interchangeably the terms groups, habitat classes, habitat types, ecosystem types, and sites. Please, unify nomenclature.

 

Done. We have standardized the term to “habitat classes” throughout the entire text. Exceptions were made in two places: at the bottom of Table 2, where the word “site” is retained because it is part of a formula and must be presented exactly as it appears in the references; a sentence has been added to the end of the table’s footnote to clarify that “Site group is the same as habitat classes.” The same exception applies to Table A2 in Appendix A for the same reason.

 

• Line 12-14. The first two sentences of the abstract do not reveal any substantial information on the study results. It can be omitted.

 

Done.

 

• Line 26-27. This statement does not follow directly from the research conducted.

 

Done. 

 

• Line 72-73. This hypothesis seems obvious; various habitats are inhabited by different assemblages of species. Especially when tree-covered ecosystems are compared to open ones.

 

We respectfully disagree with your opinion. More broadly, the same animals can be present in various habitat types as long as food resources are available, regardless of whether these environments are open or closed (tree-covered). In the case of edaphic millipedes, this idea is reinforced because they are primarily soil-dependent, and their populations usually do not vary much between ecosystems within the same region. Additionally, the habitats selected in our study are geographically close.

Please note that the phrase "the diversity of millipedes differs significantly" implies a statistical criterion: even if a difference exists, it is not necessarily statistically significant. Moreover, we are seeking indicator species specific to each ecosystem or habitat class. Some species have abundance and frequency patterns that qualify them as indicators of the different ecosystems compared. Two or more sites or habitat classes may differ in species composition, but that does not necessarily mean that any of them contain indicator species according to the INDVAL index we used.

Finally, to avoid misunderstandings, we have decided to remove the hypotheses from the article.

 

• Line 73-75. Statistical correlation is a tool in scientific research; the aim of the study is to find biological meaning using various methods, not the mathematical relation.

 

Science is and always will be an approximation to understanding reality, and there will always be different tools and methodologies to address the questions we ask. Every scientific article, even the best, necessarily has a narrow focus due to the wide variety of available methodological approaches.

In our case, this article is based on cross-sectional and non-manipulative research with a descriptive and correlational scope—in fact, these are the exact terms we use in the manuscript. We have identified correlations between millipede richness and population density and various soil physicochemical variables, determining whether these correlations are statistically significant.

This was the objective and scope of this part of the study; other research could contribute by employing manipulative or experimental methodologies. We do not make causal inferences, nor did we claim to do so in our objectives.

What matters most is to assess whether the results obtained align with and are supported by the methodology used, given the study’s stated objectives.

 

• Line 90-91. It is not clear in what sense the term connectivity is used. Do authors mean connectivity in landscape, similarity of assemblages, or something else?

We clarified the sentence by removing the word “connectivity” (line 92).

 

• Methods section. It is not clear whether the samples of Millipedes were taken three times(November 21–24, 2020; December 7–8, 2020; and January 18-22, 2021) from the same 1x1 m quadrats. If so, I wonder how previous examinations influence the diversity and abundance in subsequent ones.

We also clarified the sentence by adding “only once in each sampling unit” at the end of the paragraph (line 130).

 

• Line 185. Are you sure it's p greater than or equal to 0.05?

To determine whether there was a correlation between each soil variable and the population density and species richness of millipedes (analyzed separately), we applied the Shapiro-Wilk normality test. This allowed us to decide whether to use the Pearson or Spearman correlation coefficient in each case. If the p-value was ≥ 0.05, we used the Pearson coefficient; if the p-value was < 0.05, we used the Spearman coefficient.

We would like to add that, in this approach, we followed the more rigorous but also more realistic path in terms of results, rather than arbitrarily choosing one coefficient or transforming all data sets to apply the same coefficient. Proper statistical practice requires performing a normality test for each pair of variables and selecting the appropriate correlation method accordingly. This is the conventional procedure widely recognized in the scientific literature.

 

• Line 192-197 and 205-213. It is clear but whether the diversity indices were calculated for the sample, the transect, or the site type. The procedure of Hutcheson and the Satterthwaite approximation should be described in more detail, or the reference should be given, since it is not a widely known procedure.

Done. Diversity indices were calculated for each habitat class (site type), and this has been clarified in the text. The citations for this procedure were already included, but we have repeated them in that section for clarity. The original analysis appears in the Supplementary Materials submitted to the editor and is cited in the manuscript (see the Indexes and Tests sheet), where the full procedure is detailed.

In any case, we have added a brief description of this procedure in the main text. We took this specific observation as an opportunity to clarify the methodology section (2.3.) as much as possible, making the necessary corrections and clarifications. These changes are recorded in the change control.

For example, the term ANOVA appeared where it should have been written out as analysis of variance, an error arising from translation and repeated twice. A formal ANOVA was not applied to compare habitats. Instead, an analysis of variance of species means between the compared habitats was performed using INFOSTAT, along with the analysis of variance integrated into the weighted Shannon index hypothesis test to determine whether statistically significant differences existed between each pair of habitats.

 

• Line 222-230. There is an inconsistency in the way the results are presented, for TOV Pinoc and TUV Boslat - min, average, SE; for TOV and TUV Sabapa - max, average, SE; for BGT and RH - range, average, SE and CV. It is much better to use one scheme the data is presented.

You are right. This information appears in full in the Supplementary Materials (Table S1), and the citation is at the end of the paragraph. We wanted to highlight part of these results in the main text. To communicate this information more clearly and consistently, we have crossed out the previous wording and replaced it. The paragraph now reads as follows:

3.1. Characterization of the Sampling Transects

     The elevation of the transects was similar: 2,003–2,384 m a.s.l. (Table A1). Above vegetation (TOV) the lowest temperature value (7.5) was recorded in Pinoc, with mean and standard deviation of 13.33 ± 3.85, while the highest was in Sabapa (TOV: 25.9; 20.3 ± 4.48). Below ground vegetation (BGT) the lowest temperature was recorded in Pinoc (8; 12.1 ± 3.01), while the highest value was in Sabapa (22.5; 13.84 ± 5.24). The coefficient of variation (CV), with attention to the BGT variable, recorded the lowest value in Boslat (5.24), while the highest value corresponded to the Sabapa habitat class (37.86), followed by Ecoag (26.7). Sabapa was also the habitat with the highest variability in relative air humidity (%RH), with range of 22.6–89 (64.18 ± 25), and the highest CV in the records (38.95%), (Table S1 in Supplementary Material).

 

• Line 232 – What statistical test are the authors referring to?

This was an error. We eliminated these three lines.

 

• Line 240-247. Are the differences in Millipede’s density and species richness statistically significant? This is much more important than deliberations on data dispersion. Try ANOVA or Kruskal-Wallis test.

We completely agree with you that it is more important to explain whether or not significant differences exist. However, this preliminary description helps to further support Table 1 and Figures 3 and 4 by highlighting the most important points. Please see Table 4 and Table B1 in Appendix B, as well as the relevant sections of the text.

Table 4 presents the results of the Shannon diversity hypothesis test using the Hutcheson procedure, which includes analysis of variance (complete data are in the Supplementary Materials). Table B1 in Appendix B shows the results of the analysis of variance for 12 species across habitat classes, indicating which species show significant p-values in the corresponding habitats.

Regarding the changes you requested for Figures 3 and 4—mainly replacing the unnecessary legends with data from the descriptive statistical analysis—and substituting the following two figures with a table, the statistical analysis results will better align with the descriptive text. We have also removed some unnecessary lines detailing descriptive statistics to keep the text concise.

We have aimed to be straightforward and parsimonious in this research, in line with our objectives. ANOVA can only tell us whether there is a significant difference among the entire set of habitat classes, but it does not identify which specific pairs of habitats differ. Subsequent post-hoc tests would then be necessary, increasing the margin of error.

Moreover, the distribution of millipede data per sampling unit is not normal for at least two habitat classes, variances are not homogeneous, and sample sizes are small—thus violating three key assumptions required for ANOVA. The distribution disparity also makes the Kruskal-Wallis test unsuitable, as it requires similar shapes and dispersions across groups and is sensitive to outliers and biased values—for example, in the Sabana habitat, one sampling unit had 50 specimens while another had zero.

Besides these requirements, Kruskal-Wallis does not indicate which specific habitats differ from each other, making it as limited as ANOVA for our purposes.

For these reasons, we chose the clearest and most appropriate tools: alpha and beta diversity indices; null hypothesis testing with the weighted Shannon index to determine significant diversity differences between habitat pairs statistically rather than by simple comparison; analysis of variance of species means between habitats to identify which species and habitats show significant differences; and determination of indicator species for each habitat using the INDVAL method. These tools allowed us to achieve our objectives clearly and rigorously.

NMDS is a suitable and valid tool for our samples and has already been incorporated into the analysis following your suggestion. We also introduced Principal Component Analysis and Multidimensional Scaling.

Thank you very much for your valuable recommendation!

 

• Figures 3 and 4. The legend is unnecessary, but the axes description is mandatory. Please, add the result of the statistical analyses.

Done. The two figures were edited to include axis labels and to remove unnecessary legends, which were replaced with data that directly support the figures. Please note that it is not customary to include extensive information in figure legends. However, we have included the complete and original analysis data in the Supplementary Materials, which are available to the reader.

 

• Line 275-276. It is an interpretation, not a result.

You are right. The following sentence was deleted: "Overall, the distribution was quite skewed, with specimen clustering within certain quadrats."

 

• Figure 5. The figure is difficult to follow, especially because of the lack of description of the axes. A simple table would be more readable, in my opinion.

Okay. We've replaced the figure with a table. This required removing the reference to "violin plots" from the methodology section and eliminating the associated citations. We have also inserted the citation for the new table and renumbered the subsequent tables accordingly. However, the new table could not accommodate all the data previously shown in the figure, so we decided to include only the number of specimens, minimum–maximum values, and mean. The complete information is provided in the Supplementary Materials.

 

• Figure 6. As above,  the figure is difficult to follow. A table with species in rows and habitat types in columns, with the number of each species in the table body, would give more information about shared species and their distribution among sites

We removed the figure but did not replace it with a table. After carefully reviewing the text, we found that the information the figure aimed to highlight is already implicit in the last column of Table 1. To address what was missing, we added a new paragraph near Table 1. We also removed the paragraph that originally explained the figure.

 

• Line 300-313. There are some mental shortcuts in the paragraph. What is the “quantitative distribution”?

Done

 

• The lines 300-304 are not statistically precise.

Okay, we understand you perfectly. The inaccuracy you pointed out arises because you were expecting the results of an ANOVA and a post hoc test. However, this has already been addressed earlier, and the corresponding clarifications and corrections have been made. The results we are highlighting here correspond to the analysis of species variance across habitat classes, as previously explained; see Table B1 in Appendix B, as cited. The complete data appear in the Supplementary Materials.

 

• ANOVA reveals the differences among groups, and one statistically different value from the others is enough to find the result significant. To find which values are different, a post hoc test is required.  There is no information on the post hoc test used. The abundance of species in samples often deviates from normality, and the condition of homogeneity of variance is violated. The is no test presented that the conditions for ANOVA have been met.

This was explained above and in the previous comment.

 

• Line 342-346; 350-354. Those explanations refer to the mathematical construction of the indices but have no biological meaning. With the table suggested in the remark for Fig. 6, for those who know the construction of the indices, it would be obvious without the additional explanations.

Okay. These lines were removed.

 

• Table 3. The species richness should be presented in the table.

Done. We have added S to the table. Previously, in the text and tables, we referred to species richness in its simplest form—as the number of species collected in each habitat—denoted as S. In this part of the article, we are using that number to estimate alpha diversity through the Margalef index, calculated as (S − 1) / ln(N), where N is the number of specimens in the site or habitat studied. This explains why our three habitats with seven species each do not have the same Margalef value: the number of specimens differs. For example, Ecoag has only one specimen, and its Margalef d value is 0 (since S − 1 = 6 and ln(1) = 0, the result is undefined or interpreted as 0 for practical purposes).

However, in response to your suggestion, we have now included S in the table.

 

• Line 375 -376; 379-378. As in remark for Line 342-346; 350-354, those explanations refer to the mathematical construction of the indices but have no biological meaning.

Done. These lines were deleted.

 

• Line 380-383. It is an interpretation, not a result. Furthermore, it is not clear what the ecological connectivity or connection is.

Done.

 

• Table 4. Why do authors not use ANOVA or the Kruskal-Wallis test for comparison of the Shannon diversity index, depending on the data characteristics, and then a post hoc test?

This question is widely answered and supported above.

 

• Table 5. I suggest using NMDS to show similarity relations among samples of the habitat type studied. I would reveal the similarity between habitat types and the variability of the species composition in each habitat type as well.

Done. Following your suggestion, we performed two NMDS analyses.

Thank you!

 

• Line 400-404; 411-414. It is an interpretation, not a result.

Okay. We deleted these lines from the Results section and moved the information to the Discussion

 

• Line 410, 411, 415,423,431. Those lines refer to the literature sources. In the result section, only the own results can be presented, not the literature references.

These lines were removed from the text.

 

• Line 432 -444. The approach of calculating the correlation of each environmental variable with species variables separately, in my opinion, is outdated. Modern analyses can analyze all environmental data and the abundance of each species and ordinate species, sites, and environmental variables on one graph, with statistical tests.  It allows for the finding of correlations between environmental variables, avoiding collinearity, and finding more universal patterns in the data. CCA or RDA with environmental variables selection seams good point to start.

These techniques require a large number of sites—ideally, several times greater than the number of environmental variables—to be applied correctly. The number of environmental variables should not exceed the number of habitats and, ideally, should be at least 25% fewer. In our case, we included at least 19 variables but had only eight sites. Few papers published in Ecology include 20 or more sites; many use far fewer, yet still apply these techniques, which is not methodologically sound.

Moreover, these techniques assume that the studied populations exhibit linear environmental gradients, or in some cases, environmental optima. None of this was evident in our soil data, as shown in our value tables. Additionally, the skewed distribution of the millipedes, with the presence of singletons, compromises the reliability of these analyses.

 

 

Once more, we thank you for your valuable corrections and suggestions, which have helped us improve the manuscript.

 

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

This is interesting research and should be published. I have a few comments about the research and the manuscript:

Research:
It is not entirely clear to me that the research actually contains only 32 samples. After all, the millipede sampling ran for 11 days and one sample was taken for 21 minutes (during seven hours of each day), I suspect that I understand the design described.
The method of individual collection used IN DIFFERENT BIOTOPS is very tricky. Raking through the topsoil and collecting millipedes is relatively easy in a tomato field, but very difficult in a grassland. Millipedes can hide in the root system of grasses where they are very difficult to find. On the other hand, in a field that is exposed to high temperatures (see temperatures given in the paper), one can speculate that millipedes have hidden deeper in the soil where the authors did not find them. For this reason, I believe that the methodology presented should not be used to compare millipede densities between different habitats. It does not provide data on population abundance, but on the number of "find-able" individuals, which is influenced by vegetation type. Pitfall traps or detailed analysis (heat extraction of laboratory hand sorting) of soil samples collected would provide much better data for comparison. In presented research, abundances of ca 10 millipedes per square meter are very very low.

I consider it inappropriate to test the relationship between physical and chemical soil properties and millipede diversity using correlation coefficients; diversity indices are very susceptible to change in such small samples (few species and few individuals) by adding/removing a few individuals; for example, Canonical Correspondence Analysis (CANOCO program) would be much more appropriate, in which the effect of individual predictors (soil properties) on explaining millipede abundance can be tested. CANOCO will also help to reveal correlations of individual properties (e.g. pH versus ions), thus suggesting which factors may be directly significant and which only indirectly.

 

Manuscript

The text is somewhat disorganized. For example, specific diversity values do not belong in the abstract, the last paragraph of the introduction contains methodological information (the introduction must end with the objectives of the work), the authors formulated hypotheses, but these hypotheses are not statistical in nature; they are rather assumptions. A statistical hypothesis cannot be CONFIRMED, but only REJECTED.
The last paragraphs in the results already contain an interpretation of the data and their comparison, so it is more of a discussion. The information in the discussion about the numbers of individuals obtained in other studies is irrelevant, as different methods were used.
Some descriptions are unclear (what does “slope ranges mostly between 32-64%” mean? What is 100 percent? Area? Which area?).
I do not understand Fig. 5 - the X-axis is not described, is it really just an expression of species abundance in 4 samples? Is such a complicated sophisticated graph necessary for just 4 numbers? In the IndVal calculation, it is unclear what you mean by GROUP - why should samples from one stand be called a group? Why do you suddenly state that you are testing three groups (yes, there was only one species in the field), it is confusing. 

 

Author Response

We thank the reviewer for their valuable remarks, which have helped us improve the quality of our manuscript.

 

This is interesting research and should be published. I have a few comments about the research and the manuscript:

Research:
1- It is not entirely clear to me that the research actually contains only 32 samples. After all, the millipede sampling ran for 11 days and one sample was taken for 21 minutes (during seven hours of each day), I suspect that I understand the design described.

If we count the days listed for the three sampling expeditions (excluding the earlier prospecting trips), the total is indeed 11 days. However, these are not net sampling days, nor are the working hours straightforward, as we explain below.

We are dealing with remote locations on a mountainous plateau, and the travel to the sites—over extremely poor roads—takes 4 to 5 hours. Typically, we lose half a day to a full day of potential sampling time on each trip just to reach the sites, which accounts for 3 to 4 days.

The sampling process itself also requires preparation: setting up transects with tape, measuring and marking the grids, and taking temperature, relative humidity, and other environmental estimates (some of which had been done during the earlier prospecting trips). From a fixed base camp, we travel to the sampling sites very early in the morning. Except for two sites, the locations are 30 to 40 minutes away by road. Upon arrival, we set up a quick temporary camp and begin sampling, generally from 8:00 a.m. to 3:00 p.m., by which time it typically begins to get dark.

As explained in the methodology section, sampling may be paused at times (along with the stopwatch) for various reasons. Moreover, the millipedes must be kept alive until we return to camp to ensure proper preservation; they are not immediately placed in alcohol. Once back at camp with the millipedes and soil samples, we begin a long process of handling and processing the specimens, which often extends late into the night.

In short, this is extremely labor-intensive and exhausting work. We hope this explanation answers your question satisfactorily.

 

• The method of individual collection used IN DIFFERENT BIOTOPS is very tricky. Raking through the topsoil and collecting millipedes is relatively easy in a tomato field, but very difficult in a grassland. Millipedes can hide in the root system of grasses where they are very difficult to find. On the other hand, in a field that is exposed to high temperatures (see temperatures given in the paper), one can speculate that millipedes have hidden deeper in the soil where the authors did not find them. For this reason, I believe that the methodology presented should not be used to compare millipede densities between different habitats. It does not provide data on population abundance, but on the number of "find-able" individuals, which is influenced by vegetation type. Pitfall traps or detailed analysis (heat extraction of laboratory hand sorting) of soil samples collected would provide much better data for comparison. In presented research, abundances of ca 10 millipedes per square meter are very very low.

In previous work, we sampled significant volumes of soil using traps and heat extraction. However, in our case, the results were very poor—traps were often overturned or damaged by feral pigs or wild cats. For spiders, millipedes, and other macroinvertebrates, active manual searching and collection have proven to be more efficient.

In support of this method, we emphasize that we are rigorous and consistent with the unit of effort: the same collectors are always used, and the same amount of time is allocated for each sampling unit. This refers to net collection time, as the stopwatch is paused whenever necessary. Indeed, in these mountainous areas, millipedes often burrow among the roots in the Sabapa (tussock grass) habitat, so in that specific ecosystem or habitat class, it is sometimes necessary to stop the stopwatch in order to lift the large grasses or straw, properly expose their roots, and then continue the search.

The number of millipedes per square meter collected in this study is consistent with what is typically found in the Dominican mountains—except in certain sites where Spirobolellidae tend to cluster, or in low-lying areas near agricultural plots, where large groups of Paradoxosomatidae are often found. The skewed distribution we observed is also a common pattern: for instance, 50 specimens may be found in one sampling unit while another unit in the same habitat yields none.

Regarding the sentence: 'The surface layer was first cleared using a flat-edged spade, and then an opening approximately 60 cm wide and 40 cm deep was excavated'—this refers specifically to soil sampling. We invite you to revisit the text for clarification. In the case of millipede sampling, individuals observed on the surface are collected first, as soon as sampling begins, before excavation of the soil begins.

 

• I consider it inappropriate to test the relationship between physical and chemical soil properties and millipede diversity using correlation coefficients; diversity indices are very susceptible to change in such small samples (few species and few individuals) by adding/removing a few individuals; for example, Canonical Correspondence Analysis (CANOCO program) would be much more appropriate, in which the effect of individual predictors (soil properties) on explaining millipede abundance can be tested. CANOCO will also help to reveal correlations of individual properties (e.g. pH versus ions), thus suggesting which factors may be directly significant and which only indirectly.

The response below is part of the one provided to another reviewer who made a similar observation. Canonical techniques such as CCA and RDA require a large number of sampling sites, ideally several times greater than the number of environmental variables, to be applied correctly. The number of environmental variables should not exceed the number of habitats and should ideally be at least 25% fewer. In our case, we included at least 19 variables but had only eight sites.

While a few papers published in Ecology include 20 or more sites, others apply these techniques with fewer sites; however, this is not statistically appropriate. Furthermore, these techniques assume that the studied populations exhibit linear environmental gradients—or environmental optima in some cases—which was not observed in our soil data (see table of values). In addition, the skewed distribution of the millipedes and the presence of rare species further complicate the validity of these analyses.

In the revised version of the manuscript, we have included multivariate analyses using NMDS, PCA, and Multidimensional Scaling, which are more suitable for our dataset.

 

Manuscript

• The text is somewhat disorganized.

Okay. We have made some changes to improve the manuscript’s organization.

 

• For example, specific diversity values do not belong in the abstract, the last paragraph of the introduction contains methodological information (the introduction must end with the objectives of the work),

We revised the objectives that were previously in the penultimate paragraph and placed them at the end. The paragraph that was originally at the end has been moved to the Methodology section, and the title of that section has been adjusted accordingly. However, we believe that the diversity results should still appear in the Abstract. 

 

• The authors formulated hypotheses, but these hypotheses are not statistical in nature; they are rather assumptions. A statistical hypothesis cannot be CONFIRMED, but only REJECTED.
  We have removed the formulation of these hypotheses throughout the text. They were originally presented in statistical terms, but this seemed too cumbersome and uncommon for this type of article, so we had reformulated them in a simpler form

 

• The last paragraphs in the results already contain an interpretation of the data and their comparison, so it is more of a discussion.

We removed the last three lines, but the preceding information has been retained, as it reports results.

 

• The information in the discussion about the numbers of individuals obtained in other studies is irrelevant, as different methods were used.

We made substantial changes in response to your comments.

 

• Some descriptions are unclear (what does “slope ranges mostly between 32-64%” mean? What is 100 percent? Area? Which area?).

Okay. Changed to: slope values are predominantly within the range of 32% to 64%.

 

• I do not understand Fig. 5 - the X-axis is not described, is it really just an expression of species abundance in 4 samples? Is such a complicated sophisticated graph necessary for just 4 numbers?

Done. The figure was replaced by a Table.

 

• In the IndVal calculation, it is unclear what you mean by GROUP - why should samples from one stand be called a group? Why do you suddenly state that you are testing three groups (yes, there was only one species in the field), it is confusing. 

The name of the habitat class was left alone, we eliminated the word group.

 

Once more, we thank you for your valuable corrections and suggestions, which have helped us improve the manuscript.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

I prepared my first review by sharing all my doubts about the work with the authors in order to provoke them to revise some of their views and improve the work. The authors took advantage of many of my comments, while engaging in discussion with others to explain the broader context of their solutions. Not all of these explanations are convincing to me. The role of a reviewer is to evaluate the value of the work, not to engage in scientific discussion with the authors, so I will refer to those parts of the article that still raise my objections.

I have three comments.

1. It is still unclear why these four types of environments were chosen for the study. Was the aim of the study to compare any environments, or was there a deeper idea or hypothesis behind the researcher's choice? In response to the review, the authors explain that the research was part of a larger project and that there were technical difficulties in reaching other environments (as I understand it, the transect along the road facilitated the work). In response, the authors write, “prior knowledge of millipedes and other zoological groups in the area gave us confidence in the presence of a fauna suitable for a study of this nature.” None of these reasons explains the scientific rationale for such a selection of areas. The point is to explain what the nature of this study is. In my review, I pointed out that the hypothesis is trivial. It was also obvious in light of the research conducted by the authors 10 years ago. In response, the authors simply removed the hypothesis without replacing it with any other. The work would benefit greatly if the authors were able to find a good hypothesis to test.
2. We still differ in our approach to variance analysis and the Kruskal-Wallis test. Should we conduct a series of pairwise comparisons, exposing ourselves to the risk of a type I error? Or is it better to conduct the Kruskal-Wallis test, even though some of the conditions for this test have been manipulated, and only then perform a post hoc test?
3. NMDS diagrams 5 and 6. It would be better to show the location of all samples on the diagrams, marking the environments with colors of dots or lines surrounding a given type of surface. This would provide information about the level of variability. Of course, it may also make the diagram slightly less readable. But showing only the centroids is like showing only the averages without any measure of variability.

Apart from these comments, the manuscript has been significantly improved in terms of form and scientific content.

Author Response

A- I prepared my first review by sharing all my doubts about the work with the authors in order to provoke them to revise some of their views and improve the work. The authors took advantage of many of my comments, while engaging in discussion with others to explain the broader context of their solutions. Not all of these explanations are convincing to me. The role of a reviewer is to evaluate the value of the work, not to engage in scientific discussion with the authors, so I will refer to those parts of the article that still raise my objections.

It's true, we had differences on some points in the previous round and we may still have them, but that is part of science: clarifying, debating, testing, correcting, and continually improving in the effort to get closer to the reality we seek to understand. Differences aside, we acknowledge your critical perspective and appreciate the observations and suggestions that have helped us improve the manuscript—that is what matters most.

 

I have three comments.

1. It is still unclear why these four types of environments were chosen for the study. Was the aim of the study to compare any environments, or was there a deeper idea or hypothesis behind the researcher's choice? In response to the review, the authors explain that the research was part of a larger project and that there were technical difficulties in reaching other environments (as I understand it, the transect along the road facilitated the work). In response, the authors write, “prior knowledge of millipedes and other zoological groups in the area gave us confidence in the presence of a fauna suitable for a study of this nature.” None of these reasons explains the scientific rationale for such a selection of areas. The point is to explain what the nature of this study is. In my review, I pointed out that the hypothesis is trivial. It was also obvious in light of the research conducted by the authors 10 years ago. In response, the authors simply removed the hypothesis without replacing it with any other. The work would benefit greatly if the authors were able to find a good hypothesis to test.

Regarding why these particular ecosystems or habitat types were chosen—and not others (as we add)—we already provided our justifications in the first round of review. Here are some additional comments. The PNVN (Parque Nacional Valle Nuevo) was selected because it is one of the most important protected areas in the Dominican Republic and throughout the insular Caribbean. This park faces threats that endanger its biological diversity and the integrity of its ecosystems, which is why numerous studies and conservation efforts have been carried out there. These studies include research by foreign universities (USA, Colombia, Brazil), involving doctoral and postdoctoral projects.

Within PNVN, the region that offers the greatest continuity of ecosystems along a single route is the high plateau, where the ecosystems we selected are well represented, interconnected, and extensive. In addition to these, there is also riparian vegetation, which, as we have already mentioned, would be uninformative for ecological comparisons involving diplopods. This region has climatic, orographic, and vegetative features that make it unique in the Caribbean. In fact, Harvard University, under the leadership of Professor Bryan Farrer, has institutionalized a summer camp in the area for students aiming to pursue further studies in ecology. Currently, a doctoral student is conducting her thesis on the ecophysiology of Anolis lizards in this same region, continuing the work of renowned herpetologist Martha Muñoz, who has worked on ecophysiology and evolutionary biology in these very ecosystems.

We could also mention, among others, the work of Gustavo Hormiga (George Washington University) on spiders and Richard Glor (University of Kansas) on lizards, conducted in these same ecosystems. At the same time, the high plateau of Valle Nuevo is the region that offers logistical ease of access and makes research feasible. On the other side of the plateau, many kilometers away, lies the northern region of PNVN, where two types of forests are found—Prestoea forest and Palo de Viento forest—which are not subjected to the same pressures and threats affecting the ecosystems of the high plateau. These two ecosystems are located far from any roads and can only be accessed using pack mules, with no available facilities.

In any scientific research, feasibility must necessarily be taken into account, along with the very nature of the study and the researcher’s objectives. In the case of field ecology research, feasibility becomes a decisive limiting factor. A detailed ecological characterization of these ecosystems, in biological terms and across different seasons of the year, is considered a highly important and necessary task.

Additionally, addressing another aspect of your questions, our research includes both types of comparisons: the communities of organisms across the four habitat types, and the habitats themselves in terms of their edaphic organism communities and vegetation types.

As for the general hypotheses in the manuscript, they were indeed removed after we considered the criticisms made during the first round of revisions. However, before making this decision, we addressed the criticism that the hypotheses were supposedly irrelevant—an argument based on the erroneous idea that the fauna of two different ecosystems or habitat types must necessarily differ.

On the other hand, we accepted two very valid observations from another reviewer. The first was that the hypotheses were not formulated in statistical terms. In fact, they had been written that way in the manuscript before we submitted it to Ecologies, but two experts who reviewed the manuscript prior to submission advised us to remove them, since the use of formal hypotheses has largely fallen out of favor in many journals. For this reason, we modified them into the form in which you ultimately read them.

The second criticism from that reviewer was that in two parts of the text, we stated that some results confirmed or validated the hypotheses—criticism we accepted, as this clearly contradicted the widely accepted and well-established principle of falsifiability. Given this situation, we were left with two options: either eliminate the hypotheses from the manuscript or retain them in a statistical formulation, as we originally had them before submission. We decided on the former.

 

2. We still differ in our approach to variance analysis and the Kruskal-Wallis test. Should we conduct a series of pairwise comparisons, exposing ourselves to the risk of a type I error? Or is it better to conduct the Kruskal-Wallis test, even though some of the conditions for this test have been manipulated, and only then perform a post hoc test?

Likewise, we refer back to our response on this matter in the previous round of review. The scores in our results with diplopods turned out to be skewed, as can be verified in the main content of the manuscript and in the matrices included in the Supplementary Material. This represents one of the main acknowledged limitations that would cast doubt on the application of a Kruskal-Wallis test. However, we must admit that we are not experts in statistics—this is not our area of specialization—and you may well be right.

We will simply add one new observation prompted by a point you raise in this round. The risk of committing a Type I error is always present in any statistical analysis, which is precisely why a confidence level is established to control this type of error. In our case, we have set α = 0.05 in all analyses, meaning that we, as authors, are accepting a 5% probability of making a Type I error—this is our control.

 

3. NMDS diagrams 5 and 6. It would be better to show the location of all samples on the diagrams, marking the environments with colors of dots or lines surrounding a given type of surface. This would provide information about the level of variability. Of course, it may also make the diagram slightly less readable. But showing only the centroids is like showing only the averages without any measure of variability.

The inclusion of this non-parametric multivariate ordination analysis (NMDS) in the article is the result of your suggestions (and those of another reviewer). We chose this method because it is well suited for non-linear data and datasets with many zeros, as in our case. One of its major advantages is precisely the simplicity with which it represents the composition of ecological data in a reduced space, capturing trends or gradients while eliminating rare data points and zeros.

A common issue with multivariate analysis diagrams and graphs published in journals is exactly what you pointed out: many are not legible. For this reason, they lose their value as a means of presenting results. Introducing a high number of variables makes clarity difficult to achieve. In our first analysis (Fig. 5), there would be 32 data points for each of the six families (F1–F6) if all scores were included. In the second analysis (Fig. 6), this would simply be unmanageable, given that there are 19 soil variables. Attempting to include all the data in a single graph just to show variability would mean abandoning the NMDS approach and using a different type of graph—which, in this context, would be of little value.

We offer this preamble to properly contextualize what we did in response to your suggestion:

- Figure 5 was edited to include, in parentheses next to each vector, data on abundance, variability, and distribution (see the new Fig. 5).

- A new NMDS analysis was performed using only the three families with the highest scores (F1, F6, and F5), resulting in a new graph, which is now Figure 6 (and subsequent figures were renumbered). See new Figure 6.

- The former Figure 6, now Figure 7, was left unchanged for the reasons mentioned above. Additionally, we would like to note that this figure presents sufficient information and is complemented by explanations in the text. Moreover, our new Figure 7—now Figure 8—based on a PCA multidimensional scaling analysis, clearly presents the interrelationships among all soil physicochemical variables across the three principal dimensions of the PCA, which together account for 81.03% of the data variability.

 

Apart from these comments, the manuscript has been significantly improved in terms of form and scientific content.

We appreciate your comments and suggestions. You have contributed to the improvement of the manuscript's quality.

Reviewer 3 Report

Comments and Suggestions for Authors

Dear authors,
I disagree with your defense of the chosen method of sampling millipedes. I still believe that taking a defined soil sample (area and depth) and then conducting a thorough hand sorting without time constraints would be more reliable than attempting to quantify the time spent examining diverse areas. However, this is your research, so I respect your reasoning. Nevertheless, the abundances obtained are generally very low.

As far as the analysis is concerned, the recent procedure is much more suitable, and I am satisfied with it. However, your explanation of why CANOCO cannot be used is incorrect. CANOCO allows you to test the "simple effect" of each variable on the species distribution pattern. It can also suggest a model that uses "all" the specified variables and calculates the "conditional effect" for each of them in the given model. (This is the approach you are concerned about - many environmental variables used to explain the distribution of a small number of species.) However, at this moment, it is usually clear which variables are useless for the model. You can, of course, omit these variables and build a model with a few variables that actually contribute to its explanatory power. Those "simple effects" tests provide answers to the questions you tried to resolve using correlations in the first version. For each of them, you can then visualize the GAM. I understand that you are concerned that this may be too simplistic (which is why I do not recommend GLM), but let's be honest, your correlation coefficients are (were) even more rough and misleading.

Author Response

• Dear authors,
  I disagree with your defense of the chosen method of sampling millipedes. I still believe that taking a defined soil sample (area and depth) and then conducting a thorough hand sorting without time constraints would be more reliable than attempting to quantify the time spent examining diverse areas. However, this is your research, so I respect your reasoning. Nevertheless, the abundances obtained are generally very low.

We apologize if we did not express ourselves clearly—it was never our intention to criticize the sampling method you advocate. We have also used it successfully, but under different conditions and with objectives limited to the taxonomy of myriapods, not for quantitative comparisons. In those cases, we transported the samples to a workspace with proper lighting and other suitable conditions.

For quantitative comparisons, we have sometimes used the standard method of the Tropical Soil Biology and Fertility Programme (TSBF), employing monoliths of 25x25x30 cm. However, this method has presented some limitations when placing the monoliths in the soil, requiring us to improvise modifications that consume significant time and effort.

The method you prefer could yield results in comparative ecological studies if the sample volume to be extracted is carefully defined (as you suggest), although there is no guarantee that the number of diplopods would always be higher. Its main advantage would indeed lie in the capture or observation of the smallest diplopods, which might go unnoticed with time-limited sampling.

In the specific case of this research in the PNVN, we would not have been able to use that method anyway—unless we completed the entire sample processing on site—since removing soil from the field (even within the park) is not permitted due to its status as a protected area with strict regulations and oversight.

Thank you very much for your suggestions!

2- As far as the analysis is concerned, the recent procedure is much more suitable, and I am satisfied with it.

We are glad to hear that and grateful for your critiques, comments, and suggestions. They have helped us improve the quality of the manuscript.

 

3- However, your explanation of why CANOCO cannot be used is incorrect. CANOCO allows you to test the "simple effect" of each variable on the species distribution pattern. It can also suggest a model that uses "all" the specified variables and calculates the "conditional effect" for each of them in the given model. (This is the approach you are concerned about - many environmental variables used to explain the distribution of a small number of species.) However, at this moment, it is usually clear which variables are useless for the model. You can, of course, omit these variables and build a model with a few variables that actually contribute to its explanatory power. Those "simple effects" tests provide answers to the questions you tried to resolve using correlations in the first version. For each of them, you can then visualize the GAM. I understand that you are concerned that this may be too simplistic (which is why I do not recommend GLM), but let's be honest, your correlation coefficients are (were) even more rough and misleading.

The suggestion to limit the number of soil variables in relation to our sample size seems logical to us. We will take it into account in a future publication. Thank you again!