Here the authors present a comprehensive but in parts difficult to follow study, focusing on how local topographic and regional edaphic variation across the Amazon basin affects three facets of forest biodiversity: richness, trait structure and species composition. Using a large network of 1ha plots, the authors fit generalized least squares (GLS) models to predict each biodiversity facet as a function of local drainage topography, regional climate and geologic region. They found that drainage topography (HAND) had divergent effects on richness and species composition when plots were divided into wet vs dry regions, and that overall water deficit (MCWD) was related positively to richness- with higher diversity in well-drained soils. Similarly, the effect of drainage topography on wood density differed among geological regions within the basin. Finally, based on the results of the GLS model, the authors conclude that in plots with well-drained topography converge in species composition, whereas those with waterlogged topography are divergent in species composition. The authors conclude overall that local topography is an important factor shaping the biodiversity and resilience of Amazonian forests.

While the study was comprehensive and came to some general conclusions that were mostly supported by the data, there are three major and also some minor points that I would like to raise. The first is that the novelty statement of the paper that "the impacts of hydro-topographic conditions on tree diversity and composition remain poorly understood, especially at the whole Amazon basin scale" sounds a bit too strong and could be softened to recognize the work that has already been done. For example the authors cite several papers by Higgins et al, Moulatlet et al, Quesada et al and Sousa et al that appear to examine these topics in one way or another- what is missing from these studies? Also, in the introduction and discussion the justification for studying hydro-topography to conserve Amazonian forests in the face of climate change could be emphasized a bit more. For example in the discussion, what specifically can we use from this study to inform conservation or planning efforts? For example should we prioritize topographically diverse regions when making conservation areas?

Second, I do not fully understand how the data justify the authors assertion that plots with well-drained topography converge in species composition, whereas those with waterlogged topography are divergent in composition. I believe this is referring to Figure 2D, where the first ordination axis of plot species composition is plotted against the drainage topography variable (HAND). When I look at the plot I don't see major differences in variance across the x-axis, such that it is narrow at high HAND values and wide at low values. Perhaps there is a statical test that could be done to see if the variance in species composition changes significantly as the HAND values change.

Third, in Figure 2, how much variance does each dependent variable explain? To my eyes the effect sizes appear somewhat weak, but the lines are also sometimes obscuring each other so it's a bit hard to tell. Perhaps R-squared values or effect sizes for each line could be added to the plot or put in a table in the supplement. Either way, the magnitude of these effects should be quantified and discussed at some point in the text.

A few smaller points:

1. What was the justification for choosing the cutoff values for classifying forests as either wetter or drier (blue vs red colors in Figure 2)? To me it seems somewhat arbitrary and I can imagine the results and inference could change if different values were chosen.

2. I am wondering about collinearity in your variables. I know you tried to avoid that for the soil texture variables, but for example what is the correlation between HAND and MCWD? Between % clay and soil fertility? If possible could you provide a correlation matrix of all predictor variables?

3. Line 587: It may be worth considering redoing the analyses without the plots that followed an elevational isocline or better justify their inclusion. Such plots might differ in their topography from a rectangular plot in a systematic way and might bias the results.

4. Line 595: You mention possible inaccuracies in geographic location, I assume for some subset of plots. If the inaccuracies are large of course this could lead to erroneous results. Do you have an idea of how many plots might have these inaccuracies and how large the geographic inaccuracies might be? Meters? Kilometers?

The authors have done an admirable job revising the manuscript, however there are still some changes that could be made to improve the paper.

Major points:

Most importantly, I am a bit concerned about the switch in the modeling approach and revised manuscript (ms) from a focus on the species composition of the entire Amazon basin to comparisons of different geomorphological regions (Figure 4, Tables S1-3 in the revised ms). This is a rather large shift in focus and makes the paper somewhat less general and a bit more complex in my view, which is not ideal. I understand the authors did this to reduce species pool effects etc., but I imagine readers would still be interested in knowing how the variables of interest in this study influence species composition basin-wide, as was the focus of the original manuscript.

I wonder if it’s possible to shift the focus back to the more general basin-wide analysis of species composition in the main text, or at least add back in some basin-wide analyses to this new set of regional analyses. I am aware that Figure S4 in the revised ms covers the whole basin as in the original ms, however the assertion about species convergence towards more well-drained soils made in the original ms seems to be removed from this revision. Whether or not the aspect of convergence is re-examined in a further revision, it could be good to add back in the basin-wide analysis of effect sizes of variables of interest on species composition, as in Figure 4 D&E; of the original ms.

Regarding the novelty statement in the abstract, which in the revised ms is “the impacts of local hydrological conditions on tree diversity and composition remain poorly understood, especially at the whole Amazon basin scale,” I still think this could be softened a bit. The first half of the sentence I think is somewhat of an overstatement given the references cited in the ms and in the response letter. I think it would be more accurate to just make a small change in the text to “the impacts of local hydrological conditions on tree diversity and composition remain poorly understood at the whole Amazon basin scale” or similar. This basin-wide question is the crux of the paper and the small suggested edit makes it a bit clearer what the main goal is (but see my comment above about the shift in focus to the regional scale).

Minor points:

Thank you for clarifying that the median value was chosen as the cutoff dividing wetter from drier forests in what is now Figure 3 and Figure S2. However, in the legend for these figures I am somewhat confused as to why you then calculated and plotted the median value of each of the two groups as a label. I don’t see how this information is relevant to the paper and it could be removed to improve the clarity of the legend. The blue and red labels could then be labeled as wetter and drier instead of with the median value of the group.

Regarding collinearity of variables used in the study, it is helpful to know that the VIFs are not too high. However, rather than reporting the VIFs in a table as you do in Table S2, it would be more helpful to instead report a pairwise correlation matrix for your variables, e.g., all variables in Table S2 except region.

Regarding the plots following an elevational isoline, I am still not convinced they should be included in a study such as this about local hydro-topography. Isoline plots will by their nature of following an equal elevation contour contain less topographic variation than a rectangular plot, and I disagree with the statement on lines 480-481 that this would improve representation of hydrological conditions. As a first step could you report how many of these plots there are in the database and what percentage of the total number of plots they are?