OVERVIEW

This study explores effects of climate, soil, and local hydro-topographic variables on patterns of tropical tree species diversity, species composition, and trait composition (wood density) across Amazonia. Overall, the results show that local hydro-topographic conditions influence species diversity and community composition, but that these relationships differ among regions with contrasting climates and geology.

The study provides an important contribution to the literature on diversity-environment and composition-environment relationships at large spatial scales in tropical forests. Additional strengths of the study include an impressive number of forest plots (>400 plots) and tree species (>3,500 species), the large geographical scale of the analysis (Amazonia), and the focus on comparing the relative importance of local environmental conditions in determining tree community structure across regions with contrasting climates and geology.

The study could be improved in three general ways. First, the conceptual framework could be improved by clarifying some of the hypotheses and predictions, especially in the Introduction. In particular, I was confused as to whether Hypothesis 2 makes a-priori predictions regarding differences in species composition among regions or across environmental gradients, differences in spatial variation in species composition among regions or across environmental gradients, or both. Second, the analyses could be improved by i) clarifying which environmental variables and interactions among environmental variables are used to test different predictions, ii) providing stronger justification for using Height Above the Nearest Drainage (HAND), rather than local soil and topographic variables measured within forest plots, as a proxy of local hydro-topographic conditions in a forest plot, and iii) accounting for potential collinearity among environmental variables in the analyses. I elaborate on these comments and provide some additional suggestions below.

COMMENTS & SUGGESTIONS:

Abstract:

1. Line 6: The sentence beginning with “While these…” appears to be an incomplete sentence.

2. Line 10: If space allows, consider briefly stating why wood density was examined (e.g., what ecological insights are gained by studying patterns of wood density).

3. Line 14: Should “response” be changed to “responses”?

4. Line 15: Was community composition measured both in terms of species composition and functional-trait composition? Line 10 states “species composition”, but this result seems to refer to both species composition and trait composition.

5. Line 20: Insert “for” between “accounting” and “local”.

Introduction:

1. Line 29: Change “experience” to “experienced”

2. Line 38: Change “environment” to “environments”

3. Lines 62-75: Consider citing John et al. (2007) and Baldeck et al. (2013) in this paragraph, two studies which examined effects of local soil and topographic variables on local tree community structure in Amazonia and other tropical forests.

4. Line 75: Change “their” to “the”

5. Lines 76-78: This paragraph could be improved by highlighting the importance of dispersal limitation, along with covariation between environmental conditions and geographic distance, in determining spatial variation in community composition. The paragraph seems nicely focuses on “neutral processes”, but it is a little unclear what those processes are. One simple solution would be to change “neutral processes” to “neutral processes and dispersal limitation” in the topic sentence, and to briefly highlight the covariation between environmental conditions and geographic distance (e.g., Tuomisto et al. 2003) later in the paragraph.

References:

Baldeck C.A., et al. 2013. Soil resources and topography shape local tree community structure in tropical forests. Proceedings of the Royal Society B 280: 20122532.

John, R., et al. 2007. Soil nutrients influence spatial distributions of tropical tree species. PNAS 104: 864–869.

Tuomisto, H., K. Ruokolainen, and M. Yli-Halla. 2003. Dispersal, environment, and floristic variation of western Amazonian forests. Science 299: 241–244.

1. Lines 92-94: Add a little more explanation of how wood density relates to plant hydraulics. I am unsure what is meant by “water availability filtering”. Does this mean that water-limited environments select for individuals or species with low wood density?

2. Line 97: Change “raised” to “tested”

3. Hypothesis 1 (Lines 99-101): This hypothesis could be worded more clearly. Is the hypothesis that tree diversity should decrease with soil-water availability in wetter forests, but increase with soil-water availability in drier forests?

4. Hypothesis 2 (Lines 102-107): This hypothesis could be worded more clearly. I am unsure what is meant by “opposing patterns of species composition changes between wet and dry forests.” For example, does this refer to spatial variation in community composition between wet and dry forests (i.e., beta-diversity between wet and dry forests), or beta-diversity within wet forests compared to beta-diversity within dry forests. In addition, if the hypothesis makes directional predictions about patterns of species composition between wet and dry forests (e.g., higher beta-diversity in wet compared to dry forests), it would be helpful to state the directionality in the hypothesis. Lastly, the conceptual rationale for the second part of the hypothesis—referring to effects of soil age on community composition—does not seem to be presented earlier in the Introduction. Why would beta-diversity be expected a-priori to be higher among sites on older soils compared to among sites on younger soils?

Methods:

1. Morphospecies: Since exclusion of morphospecies is likely to remove a large number of rare species from the vegetation datasets, it would be helpful to report the number of morphospecies that were removed from data prior to analysis (e.g., the mean, minimum, and maximum number of morphospecies removed per forest plot).

In addition, removal of morphospecies could potentially bias estimates of local species diversity across the environmental (hydro-topographic) gradients of interest. To explore this potential bias, it would be helpful to examine and report how the proportion of morphospecies per plot (number of morphospecies / total number of species) varies across the gradients of interest. If this correlation is relatively weak, it would help justify exclusion of morphospecies from the analyses of local species diversity.

1. Environmental variables (Lines 137-190): It would be helpful to add a short paragraph or sentence at the beginning of this section to briefly i) summarize the number and types of environmental variables used in the analyses, and ii) clarify which environmental variables were and were not used to estimate local hydro-topographic conditions of each forest plot. For example, the first paragraph states that the Height Above the Nearest Drainage (HAND) was used to estimate local hydro-topographic conditions, but it is unclear why the two soil variables described later in this section—soil fertility and soil texture—were not used to estimate hydro-topographic conditions, especially since soil fertility and texture are direct measures of soil conditions.

The studies uses the Height Above the Nearest Drainage (HAND) as a proxy for local hydro-topographic conditions. However, it is unclear how well this variable correlates with soil or topographic conditions in the 1-ha forest plots. I would expect that soil or topographic variables have been measured in several plots in the ATDN network. If so, the authors could use those subsets of measurements to confirm that the HAND variable correlates well actual measurements of soil or topographic conditions.

In addition, it is unclear whether and how the HAND variable accounts for small-scale variation in topography within a 1-ha forest plot. For example, for two adjacent 1-ha plots with the same mean elevation but different slopes (e.g., a flat plot versus a plot located on a slope), would the HAND variable be the same?

1. Collinearity among environmental variables: The analyses do not seem to account for collinearity among environmental variables. To address this, the authors could either report the correlation between each pair of environmental variables and only use variables that weakly correlated, or perform a PCA on the environmental variables and use PC axes as predictor variables.

2. Analyses of species composition: In general, it is somewhat unclear in the Introduction and Methods as to whether the goal of the study is to test hypotheses concerning i) differences of species composition among regions, ii) differences in variation in species composition (i.e., beta-diversity) among regions, or iii) both (see my similar comment about Hypothesis 2 above).

3. Predictions of the hypotheses: It would be helpful in the Introduction or Methods to clarify some of the key predictions of the hypotheses. In the “Data analyses” section of the Methods (Lines 234-275), for example, it is unclear i) which environmental variables are being used to test each hypothesis (e.g., see my comment above about which environmental variables used to determine local hydro-topographic variables), and ii) how interaction terms in the models are being used to test predictions (e.g., the interaction between HAND and MCWD).

Results:

1. Figure 2: In the figure caption, i) define the acronyms HAND and MCWD, and ii) clarify what P-value was used as cutoff to differentiate a “significant effect” from a “non-significant effect (NS)”.

Discussion:

1. Parts of the Discussion could be improved by noting that additional analyses would be needed to explicitly test for environmental filtering. For example, environmental filtering is often tested by comparing observed community weighted mean (CWM) trait values, or variance in trait values among species in a local community, to a null model of random community assembly. In the present study, the authors simply report correlations between CWM wood-density values and environmental variables, rather than testing whether traits are more similar among species (i.e., lower trait clustering or more trait underdispersion) at certain locations along the environmental gradient.

2. The Discussion section is a little long (13 paragraphs) and could be simplified by focusing on the results most germane to the hypotheses.

I reviewed the first submission of this manuscript (Reviewer 1). Overall, the authors have nicely addressed the reviewer comments in the revised manuscript. Below I provide some minor suggestions to help improve clarity.

Abstract:

1. Line 16: Replace the acronyms for the three regions (“CA-EA, WA, SA”) with region names.

2. Lines 14-15: Given the insignificant main effect of HAND and the significant interaction between HAND and Region in Table 1, should this sentence be modified to read “Community-weighted wood density means are influenced by hydro-topographic conditions, also increasing toward better drained soils in two of the four regions.”?

Introduction:

1. Line 35: Change “world” to “world’s”

2. Line 36: Change” modulated” to “modulate”

3. Line 40: Change “As consequence,” to “As a consequence,”

4. Line 49: Change “Other” to “Another”

5. Line 75: Change “water available” to “water availability”

6. Line 78-80: This seems to be an incomplete sentence.

7. Line 81: Change “allows understanding” to “provides insights into”, and “species filtering” to “environmental filtering”

8. Lines 102 and 108: Change "species filtering" to "environmental filtering"

9. Line 114: Change “more evident” to “larger”

Methods:

1. Lines 123-125: It would be helpful to briefly state whether all 1-ha plots have the same dimensions or shapes (e.g., 100 x 100 m), or if not, the range of dimensions or shapes.

2. Lines 135-136: I don’t understand the biological justification for excluding “6 plots with less than 30 species, as they have a high abundance of a few species that may decrease the confidence in diversity and composition patterns…”. The low species diversity of those plots seems to be a real, and potentially informative, biological pattern.

3. Line 142: Add the acronym “[HAND]” after “Height Above Nearest Drainage” (i.e., to define the acronym the first time it is presented).

4. Line 143: Change “regional” to “the regional”

5. Line 167: Change “regional” to “the regional”

6. Line 183: Change “The proxy to the soil fertility was” to “We estimated soil fertility using”

7. Lines 212-213: If the CWM of wood density of each forest plot is simply calculated from the species present in a forest plot, it is unclear why it is necessary to consider the species pool in the analysis. I have the impression that the authors are using the term “species pool” to simply refer to analyses conducted within single biogeographic regions, but that the analyses themselves do not explicitly incorporate the species pool as a variable.

8. Lines 247-249: This method is a little unclear. If the goal of the analysis is to quantify variation in species composition within each region, why would that analysis use Bray-Curtis dissimilarities calculated from all plots across all regions? My understanding of this analysis is that each region would have its own centroid, and variation in species composition would be calculated using the mean or median distance to centroid of plots in the same region.

9. Line 252: Define the acronym “CWM-wd”

10. Line 254: Change “regional” to “the regional”

11. Line 269: “In Interactions” is an unclear. Should this read “To examine statistical interactions”?

12. Line 287: Change “runned” to “performed”

Results:

1. Table 1: 1) Define the acronyms “MCWD” and “HAND” in the table legend; 2) The acronym “FA”, referring to Fisher’s alpha, should be added before “Species”, “Genus”, and “Family” in the table; and 3) Line 874: Change “deal with the” to “account for”

2. Table 3: Line 878 - Change "deal with the" to "account for"

3. Figure 2: Since this figure does not show results for species composition, “species composition” should be deleted from the first sentence of the figure caption.

Discussion:

1. Line 454: Change "species filtering" to "environmental filtering"

2. Line 491: Change “climate” to “climatic”