Causes and consequences of liana infestation in Southern Amazonia
by Simone Reis et al.

The manuscript evaluates the importance of soil fertility, climate, fragmentation and host characteristics as predictors of liana infestation from tree-level to plot-level in the southeastern part of Amazonia. In addition, the influence of liana infestation on tree growth was analysed.
This study quantifies the liana crown occupancy index (COI), in contrast to other studies, which measure liana stems or liana basal area associated to individual tree stems.
This manuscript is based on a huge amount of liana data from a large area and contributes to our further understanding of liana ecology.
The presentation of the results and especially the description of the statistical methods have to be improved, but the article is overall well written and only needs some revisions to be acceptable for publication.

See my detailed comments below:

SUMMARY

Line 32: ... but not by the studied climate or soil fertility variables...
Line 44: I suggest to delete the part ", but not to climate or soil fertility". The climatic variation is limited in your study area and your soil variables are concentrations that do not necessarily represent availability of nutrients.

INTRODUCTION

Line 99f: How is pulsed deposition of nutrients related to fragmentation?

METHODS

Line 139-145: Please add the elevational range of the study plots here and the plot elevations to table S1.
Line 194-196: Base saturation is the percentage of cation exchange capacity occupied by the basic cations (Ca, Mg, K, Na).
Line 213: Were only tree species identified?
Line 249: I would include the variable "Percentage of broken trees" under forest structure.
Line 250-252: Did you use only additive GLMs or were interactions included? From the results (Fig 3, table S4) it becomes not clear how the authors selected their models to get the six resulting predictors from the initial 12 predictors (table 1). With the small number of replicates (N=27) this would be interesting to know.
Line 268: Taxon level...
Line 271: ... potential diameter growth rate...

The procedure for model selection is not clear from the methods section. For all three levels (plot, tree species, tree growth) there are several predictors given in table 1 (all included in equations 1 and 2), but the results (figures 3, 4 and 6 and tables S4, S5 and S12) only show a subset of these predictors and their importance is averaged across the best models. Instead of showing coefficients averaged across models (how many?) I would prefer to know which model combinations have been tested and then would like to see the results of the best models. I suggest to list the final models or at least the best models in the supplement.

RESULTS

Line 332-333: ... trees have low height:diameter ratios...
Line 333-337: How was the exact model structure of the mentioned models, did they include random factors?
Line 337: ...on liana infestation.
Line 341-342: This sentence should be changed: The tree individuals with larger diameters are less slender and have more light-exposed crowns (Figures S7 and S8)
Line 355-358: Percentages given in the text do not correspond to data in figure 5, there are two values for infestation above 80% and one around 20% in the figure.

DISCUSSION

Line 415-417: a positive association of liana biomass with more fertile soils was found by Fadrique & Homeier (2016) in Ecuador
Line 462-466: In Figure S8 you see that slenderness is decreasing with tree size and age. So the younger trees of each species are more slender and receive less light (Fig S7). That means that not slenderness is the real predictor of liana infestation but tree age. Tree are accumulating lianas when growing as was shown in the cited references (line 458).
Line 474-476: This conclusion is not supported wit the presented data. The more important driver changes in tree architecture and forest biomass probably is nutrient availability.
Line 504-507: reduced tree diameter growth by liana infestation was also found by Fadrique & Homeier (2016) in Ecuador
Line 520-522:You showed that physical structure and species identity (that includes wood density and slenderness, but also several additional tree properties potentially relevant for liana infestation) are more important than your extensive set environmental factors (not including nutrient availability).

FIGURES & TABLES

Figure 1: Instead of the Amazon map I would prefer a S America map on the left side of the figure.
Figure 3a:
Figure 3b: The graph contains only 19 plots. Are these the terra firme plots only?
Figure 6: Why was the analysis restricted to smaller trees with intermediate wood density?
Figure S7 and S8: Is there no statistics for these figures?
Figure S11, last two lines of legend: in parentheses it should be ">=20 individuals", names of the species are in table S10.

Table S1: Add plot elevations
Table S2: Add base saturation
Table S5: What happened to the other predictors? What amount of variance is explained by the random factors?
Table S10, first line of legend: in parentheses it should be ">=20 individuals"

The authors addressed all reviewer concerns and substantially improved the manuscript, the article is well written and, as mentioned before, will contribute to our better understanding of liana ecology.

Line 429: it should be "g cm-3" in the figure legend instead of "gm cm-3"